Eckhard Bick


PORTUGUESE SYNTAX



7. Clause types

Ordinary clauses could be regarded as a kind of group - big "verb phrases" , with a verbal head governing clause level arguments and adjuncts, but here we will treat the term vp as the form category of complex predicators (i.e. verb chains with auxiliaries, cp. chapter 5).

Verb/predicator-containing clauses can be subdivided into finite (fcl) and non-finite (icl) clauses. In an fcl, the main verb (or the first auxiliary in a vp verb chain) is finite (i.e. tense-inflected), in an icl it is not. Non-finite clauses can be infinitive clauses, gerund clauses and participle clauses.

Clauses that are constituents of a larger syntactic unit (typically another clause or a pp), are called subclauses. In Portuguese, all finite subclauses are introduced and subordinated by a so-called complementizer (a conjunction, relative or interrogative), while non-finite subclauses usually are not.

Semantically, clauses can be described as predications, where something (the predicate) is predicated of something else (the subject) . In syntactic terms, the relation between subject and predicate is called a nexus. In ordinary Portuguese clauses, the nexus-link between predicate and subject is mediated by a (verbal) predicator. Predicators are part of what is predicated, contributing between next to no content (copula verbs linking predicatives, e.g. ser, estar) or all of the content (intransitive verbs, e.g. trabalhar, dormir).

In the examples, subjects are in italics, predicates are underlined and predicators are in bold face:

Hipopótamo come muito. (transitive verb, predicator as part of predicate)

Ele era um herói nacional. (content-less copula predicating Cs predicative)

A criança dormia. (intransitive verb, predicate and predicator are identical)
 

It is a special feature of Portuguese (and most other Romance languages), that subjects are optional constituents, and can be incorporated into verbal inflexion endings. Therefore, one-word predications (a-d) or vp-predications (e), without a syntactically visible nexus, occur frequently:

(a) Chegou.

(b) Chove.

(c) Começamos!

(d) Coma!

(e) Foi vencido.

Utterances like the above fit the form categories of either word (a-d) or group (e). However, since all 5 utterances are predications and feature predicators, it is tempting to also classify them as clauses. This, however, is in conflict with the first condition in our definition of a clause as (1) a multi-constituent nexus (2) featuring a predicator and/or a subordinator. The problem gets even worse if one reads enclitic or mesoclitic pronouns not as individual words, but as morphological parts of the verb. One solution is to make a distinction between the concepts of sentence and clause. A sentence, defined simply as the top node of any syntactic analysis, does allow v-only or vp-only predications, while a clause, with its multi-constituent condition, does not. Thus, the 5 predications above are sentences, but not clauses.

In the system advocated here, hyphenated enclitics will be regarded as syntactic constituents with their own branch in the syntactic tree, and in v-only or vp-only utterances the form-tag of clause (fcl) and the function tag of predicator (P) may be used optionally, creating an additional (non-branching!) node in the analysis:

as opposed to a non-clausal analysis:

Now, predicating nexus-relations can be found in other than subject-predicator structures, too. Object predicatives (object complements, Co) or argument adverbials, for instance, can predicate something of a direct object (Oacc), not the subject - introducing a secondary nexus into the same clause, making double use of the clause's predicator.

(a) Pôs a arma no chão. (Oacc - Ao)

(b) Chamou o projeto uma desgraça. (Oacc - Co)

(c) Bebe o chá quente! (Oacc - fCo)

Sometimes, however, predications are subordinated by a complementizer (clause header) - and thus, isolated from the parent clause's predicator - without providing an additional predicator. Here the concerning predication is a nexus between explicit and implicitly anaphoric material (the latter expressed in the parent clause), linked not by a predicator but by a subordinator.

Thus, (c) can be turned into a 2-clause construction by adding a subordinator:

Here, quente is still predicated of chá, - the latter is implicitly present in the subordinated constituent, providing for an averbal nexus. We will call such subordinated verb-less constituents for averbal (sub)clauses (acl) . The acl's clause body (all but the subordinator) can be tagged with the dummy tag SUB< (argument of subordinator), but functionally it deserves a predicative tag C (or adjunct predicative, fC, since it predicates chá without a copula).

For more discussion of clausality, see chapter 6.3. on pp's (com/sem as subordinator), chapter 5 on clause hierarchies in verb chains, and chapter 4 on subordination.

symbol category examples
cl fcl, fs finite (sub)clause
oração finita
finit (led)sætning
Não acredito que seja verdade
  icl, is non-finite (sub)clause
oração infinita
infinit (led)sætning
Consertar um relógio não pode ser fácil
  acl, as averbal (sub)clause
oração averbal
averbal (led)sætning
Ajudou onde possível

7.1. Finite subclauses

Finite subclauses cover a wide range of constituent functions. Most "cognitive" verbs, for instance, allow or even demand a que-clause (a) or a finite interrogative subclause (b) as direct object:

Using a traditional - word class analogous - typology, one can distinguish between finite subclauses that substitute for nouns (nominal subclauses), adjectives (attributive subclauses) or adverbs (adverbial subclauses), respectively:

7.1.1. Nominal finite subclauses (S:fcl, O:fcl, C:fcl, DP:fcl)

with absolute relative pronoun or adverb:

Quem cedo madruga .... (S)

Molesta quem aparecer. (Oacc)

Seja quem for (Cs)

Mostrava a pedra a quem quisesse ver. (DP)

O pai não veio para o aniversário dele, o que não o surpreendeu. (fCsta)

A proposta de que ele venha para aqui não me parece realista. (DP)

with interrogative pronoun or adverb:

Perguntou quem lhe mandaria o presente. (Oacc)

Não sei quando ele chegou. (Oacc)

With the completive conjunction que:

Soube que foi o único candidato. (Oacc)

Só foi avisado depois que o seu jatinho levantou vôo. (DAarg/DP)

Nem lhe parece estranho que o Pedro tenha comprado o sítio. (S)

Levou o projeto ao fim sem que ninguém lhe ajudasse. (DP)

7.1.2. Attributive finite subclauses (DN:fcl)

modifier function, with postnominal relative pronoun or adverb:

O homem que encontrei ontem (DN)

A amiga com a qual apareceu na festa (DN)

O ano quando se casaram ... (DN)

Note that the relative clauses in these examples are all restrictive, which is why there is no comma. So-called parenthetic relative clauses are surrounded by commas, the difference corresponding to the difference - in our terminology - between ordinary adnominal modifier (DN) and predicative adnominal (DNc):

O professor, que já não suportava o calor, terminou a aula cedo. (DNc)

argument function, with se or interrogative pronoun/adverb:

Não há informações se vão levar um proceso contra o coronel. (DNc)

7.1.3. Adverbial finite subclauses (A:fcl)

adjunct function, with relative adverbial or subordinating conjunction:

João não fez nada para que ela voltasse. (fA, purpose)

Entraram na vila quando amanheceu. (fA, time)

Desliga, amor, que tem gente na linha! (fA, cause)

Faz como quiseres! (fA, manner)

argument function, with relative adverb:

Meu avô mora onde o mato começa. (As, argument adverbial)

7.2. Non-finite subclauses (icl)

7.2.1. Infinitive subclauses

Infinitives make up for the "noun"-morphology of verbs. So infinitive-icl's are primarily used where nouns would be used, as subjects, objects, complements and arguments of prepositions, i.e. as what we in ch. 7.1.1 have called nominal subclauses.

Infinitive clauses as clause level argument:

Retomar o controle foi difícil. (S)

Manda o filho comprar leite. (Oacc, causative)

Viu o marido bater na mulher. (Oacc, perceptive "ACI")

Julgo o carro ser caro demais. (Oacc)

Não temos onde morar. (Oacc)

O problema é não sermos bastante fortes. (Cs)

Chama isso fazer tábua rasa. (Co)

Infinitive clauses as argument in pp

Era uma proposta difícil a entender. (DP)

Para lhe ajudar, propôs outra solução. (DP)

Para o amigo lhe ajudar, bastava uma palavra só. (DP)

Pede para você ficar com ele. (DP)

A idéia de comprar outro carro não foi nova (DP)

Levantou a pergunta de onde abrir banca. (DP)

Both infinitives and que-clauses are very rare as direct arguments of nous, and might simply be cases where a preposition has been "forgotten" (de in the last two examples).

Infinitive clauses as adjunct adverbial

Veio lhe agradecer pessoalmente. (fA)

Foi à televisão recitar o documento. (fA)

This construction is restricted to movement verbs with a valency allowing (direction) argument adverbials: ir, vir, correr, apressar-se etc.

7.2.2. Gerund subclauses

Gerunds account for the "adverb"-morphology of verbs. So gerund-icl's are primarily used where adverbs would be used, i.e. as adverbials (A) . Most common are adverbial adjuncts (1), while argument function is rare (2), apart from a special construction with ter/haver (3):

(1) Gerund clauses as adjunct adverbial:

O novo governo acabou com a política isolacionista do anterior, abrindo o mercado brasileiro para empresas multinacionais. (fA)

Falando do João, não quero convidá-lo. (fA)
 

(2) Gerund clauses as clause level argument ("accusative with gerund")

(2a) Como imaginá-lo partilhando à vera a administração com outros? (Oacc)

(2b) Como imaginá-lo partilhando à vera a administração com outros? (Co)

Argument gerund-icl's are restricted to so-called matrix verbs with a respective valency, reminiscent of the ACI-constructions discussed in chapter 7.4. Here, too, the "accusative" (lo) can either be regarded as direct object of the main clause verb, or as subject of the subclause (gerund) verb. The latter reading yields an Oacc:icl-reading (2b) with a surface subject (the accusative pronoun) within the subclause, while the other provides for a Co:icl-reading (2a) of a gerund clause with no surface subject:

(3) Gerund as argument of ter/haver:

Tem gente morrendo de fome no Brasil. (Oacc)

Tem o motorista esperando. (Oacc)

Sentences like these can be analysed as "accusative with gerund" constructions, too:

Especially in the second case, the gerund could also be read as a free object complement, as in the sentence Tem um amigo na casa: Tem o motorista esperando. Tem um amigo na casa

Finally, though not advocated here, the small gerund constituent of the last analysis could be seen as an - attributive - DN-dependent of a larger direct object np, as one would in the case of a relative clause ("gente que morre de fome no Brasil"):

(4) Gerund with prepositional "complementizers":

Another (fixed) clausal gerund construction occurs with the preposition com and sem. These two prepositions can function as a kind of "complementizer" in creating clause-like adverbials where the np that would ordinarily be the nominal argument of the preposition (DP), is made ("subject"-) part of a clausal nexus by gerund- pp- or ap- predicatives:

Lançaram a novela com um ator étnico estrelando. (DP)

Com a Guarda Civil patrulhando a cidade, não tinha onde se esconder. (DP)

In a "DP with gerund reading", like in the "accusative with gerund" construction (2), we seem to have two analysis options of different depths, one where the DP is regarded as one complete icl, and another one, where the gerund clause is smaller and read as a predicative. This would turn the DP into an acl with a subject and a subject complement (Cs or As), the latter consisting of a gerund icl. However, the "big icl" analysis (4a) is "flatter" and simpler than the acl analysis, and also supported by the fact that the gerund alternates with a+INF-constructions the same way gerund and a+INF alternate after estar - suggesting estar as missing [zero-constituent] auxiliary of an ordinary predicator: Com a Guarda Civil a patrulhar a cidade, ....

(4a)

A third reading, that of a postnominal gerund clause (DN:icl) is ruled out by a substituion test: *Com a Guarda Civil, não tinha onde esconder-se has a completely different meaning, and therefore, the gerund-icl cannot be part of an np with [a] Guarda Civil as head.

Finally, the preposition can be regarded as a subordinator in an even larger gerund-icl, amounting to a flat clausal analysis for the whole adverbial, without a pp- or DP-constituent. The prp-subordinated gerund-icl is consistent with similar analyses for other (non-gerund) clause bodies (cp. Chapter 6.3):

(4b)

7.2.3. Participle subclauses

Participles are the "adjective" variant of verbal morphology. In Portuguese, only past participles are productive, the original present participle endings -ante, -ente and -inte having been degraded to affix status. Attributively used past participles (-ado and -ido) are inflected for gender and number, like adjectives, and the prototypical (inflecting) participle-constructions occur, like adjectives, primarily as postnominal modifiers (DN) and predicatives (Cs, Co, fC, DNc) . Another, "verbal", use of participles is in verb chains after ter (expressing tense), where there is no inflection. Finally, participles occur in ablativus absolutus constructions as pivot of a type of adverbial subclause.

7.2.3.1. Attributive participles

Attributive participles can completely turn into adjectives, and form ap's taking intensifier modifiers. Dictionaries usually list these participles individually as adjectives, and if used without heavy pp-dependents, ap-analyses are just fine (cp. chapter 6.2):

However, if more dependents - or even arguments - are added, an icl-analysis seems more and more natural. One advantage is, that the parent-verb's valency structure - and with it, clause level dependent terminology - can be borrowed.

(a)

(b)

7.2.3.2. Participles in verb chains

Uninflected past participles are used after ter to form the perfeito composto and mais-que-perfeito composto tenses:

Participles also occur in two types of passive verb chains, "action passive" (after ser) and "state passive" (after estar). In both cases the participle has inflection agreement with the subject. Participles in action passives (a) are more verb-like, an agent of passive (the original subject in the active clause) can be added (fApass), and the participles cannot be modified by DA-only-modifiers like nada. Participles in state passives (b) can be modified by DA-only-modifiers like nada, and adding an agent of passive seems odd. Therefore, we will tag participles in action-passives as main verb (MV:v-pcp) in a complex predicator (P:vp), while participles in state-passives will be assigned the "adjectival" function of subject complement (Cs:v-pcp) , or - if part of a group - head function in a Cs:ap.

A verb's valency determines in which verb chains its participle can be used. Transitive verbs with agent subjects and patient objects (abrir, fechar, comer) can be used both after ter, ser and estar, ergative verbs with patient subjects without objects (chegar, desaparecer, nascer) only after ter and estar, and intransitive verbs with agent subjects without objects (trabalhar, rir, brincar) only after ter. A semantic explanation is that verbs without a patient-argument (inergative intransitives) logically can't form any passive, while only verbs with both a patient- and an agent-argument (transitives) can form action-passives.

  ter + MV:v-pcp estar + Cs:v-pcp ser + MV:v-pcp
transitive

S-agent + O-patient

+
+
+
ergative

S-patient

+
+
 
intransitive

S-agent

+
   

7.2.3.3. Ablativus absolutus

Ablativus absolutus (the term used for Latin) or absolute participle constructions are adjunct adverbial clauses featuring an inflecting past participle as predicator and a patient subject. Like in state-passives (cp. 7.2.3.2), only transitive and ergative verbs qualify for this construction, since only they have patient arguments. If the ablativus absolutus is "unfolded" into a finite active clause, its patient subject becomes direct object (Oacc) for transitive verbs (a), but remains subject (S) for ergative verbs (b).

(a)

arrancada a chave (S) da vítima, sumiu na mata

--> arrancou a chave (Oacc) da vítima, e sumiu na mata
 

(b)

sumido o bandido (S) na mata, as vítimas se consolaram

--> o bandido (S) sumiu na mata, e as vítimas se consolaram

Absolute participles must not be confused with sentence initial participle ap's (or - if preferred - participle clauses). The latter function as adjunct predicatives (fC) , the former as adjunct adverbials (fA). Predicative participle clauses (b) have the same subject as the main clause, and inflect accordingly. Absolute participle clauses (a) have their own explicit subject, and no agreement with the subject of the main clause.

(c)

(d)

7.3. Averbal clauses (acl)

In our terminology of Portuguese syntax, averbal (sub)clauses consist of a subordinator (or complementizer) and a clause body featuring one or more clause level constituents, but no predicator (which would make the clause an fcl or icl). As a dummy function tag for the clause body, the tag SUB< (subordinator argument) is used.

In the example (a), the subordinator is a relative adverb, which has its own in-clause function (fA) on top of the subordinator function - which is why no SUB-tag is used. Conjunctions, on the other hand, are pure subordinators:

In most cases, the acl clause-body has predicative function, and in (a/b) SUB< could be replaced by a more functional Cs (quando [está] doente, embora [esteja] doente), or - if one doesn't want to think of a zero-constituent copula - fC (adjunct predicative).

It is due to this "predicativeness" that the participle in (a3) is not treated as a predicator, but analogously with doente (a1) and criança (a2). Cp. also the predicative participle discussion in chapter 7.2.3.4.

Gerunds, too, can be used predicatively in acl's:

Using the acl-analysis, and not counting morrendo/cansado as predicators, is a useful way of distinguishing between (c) and (c'), which would otherwise receive the same analysis (P & icl):

Some relatives (como, quanto, qual) and the conjunctions que and do_que can function as comparative subordinators (SUBcom) and introduce comparative acl's, both as clause (d1) and group (d2) constituents.

In both acl's, the clause body tag SUB< has been replaced by more functional tags, fA (adjunct adverbial) in (d1) and S (subject) in (d2), respectively. Such tags are, however, controversial, since they depend on which "zero constituents" and thus, on what kind of "unfolded" clause structure one imagines.

With a third kind of subordinator, prepositions, the acl analysis also suggests itself as an alternative solution for those special pp's - headed by com or sem - that contain predications (discussed in chapter 6.3.):

Finally, acl's may come handy in the top-level analysis of certain averbal utterances that nevertheless feature a kind of clausal nexus. So far, we have been discussing subordinated averbal clauses only, and here - in Portuguese - complementizers (conjunctions, relatives or prepositions) are obligatory, as we assumed in our definition of acl's. But what about averbal main clauses, as they occur in, for instance, exclamations and headlines:

7.4. ACI and causatives (transobjective constructions)

Like other Romance or Germanic languages, Portuguese features some so-called matrix verbs that govern infinitive-subclauses with independent subjects (i.e. different from the matrix clause's) that take not nominative case, but accusative (or, in some cases, dative) form. Thus, what semantically represents the subclause subject is morphogically marked as object in the main clause (matrix clause). This can syntactically be interpreted in two obvious ways: (1) Either the surface constituent in question is really subject of the subclause, but bears a morphological case mark for the whole (object) subclause. (2) Or the surface constituent is really object of the matrix clause, and the subclause functions as object complement, with its own subject anaphoric and unexpressed at surface level.

In order to force case on Portuguese nominal constituents, pronoun substitution is useful, as in the following examples of infinitive clause candidates for direct object function:

  • (a) Não deve contar isso. Não o deve. (auxiliary with AUX< complement)
  • (b) Julgo (eles) serem inocentes. (main verb with Oacc argument)
  • (c) Vi-o bater na mulher. (sense-verb with ACI construction)
  • (d) Fizeram-nos trabalhar nas minas. (causative construction with accusative)
  • (e) Permitiu-lhe usar o nome da empresa. (causative construction with dative)
  • (a) is the prototypical auxiliary complement case, with the two verbs' subjects coinciding and - as a verb chain test - pronoun fronting of the second verb's object left of the first verb: Não o deve contar. (b) is the typical main verb case, with an independent nominative subject in a direct object subclause. The difference can be shown either in form (vp vs. icl constituent), or in function (AUX vs. MV and AUX< vs. Oacc function), as described in the chapter on verbal constituents (ch. 5).

    The transobjective construction in (c-e) are more problematic: The two verb's subjects differ, and the second verb's object can't be pronoun-fronted (*a vi-o bater), suggesting an analysis with two main verbs, as in (b). On the other hand, the second verb's subject is marked as object of the first, it is hyphen-linked to the "wrong" verb, and even "frontable" (o vi bater na mulher).

    (c) is what in Latin is known as accusative cum infinitive (ACI), and is restricted to sense-verbs: ver, ouvir, sentir. (d) is called a causative construction: X causes (Y do/happen). Also causatives constitute a restricted class: fazer, deixar, mandar. Analysing (c) and (d) the same way as (b), we get:

    An argument in favor of the Oacc:icl constituent is that it can be replaced by an ordinary Oacc:fcl, like in the non-matrix-cases (b): Vi que ele batia na mulher. Fizeram que eles trabalhassem nas minas. This does, however, involve a change from accusative to nominative case for the pronoun, and we could try another analysis, that gives full (syntactic) credit to morphological form:

    In this analysis, the accusative pronoun functions as direct object in the matrix clause, and the subclause functions as object complement.

    In similar cases, with a pp or ap object complement governed by a sense-verb or a causative verb, this second analysis has the additional advantage of not needing to introduce an averbal subclause without a complementizer or predicator. Compare:

    (1) Object complement analysis:

    (2) Averbal clause analysis:

    On the other hand, there seems to be a syntactic difference between ACI's (c) and causatives (d) in that the object subclause can be substituted by the accusative pronoun alone in (c), but not in (d) - though even in (c) this is semantically problematic.

    (c') Vi-a bater no marido. - Vi-a.

    (d') Fizeram/deixaram-nos trabalhar nas minas. - *Fizeram/deixaram-nos.

    This suggests that the Co:icl analysis fits ACI's better, since it assigns the pronoun Oacc function in the first place - while the Oacc:icl analysis yields a better fit for causatives. If I see somebody hit her husband this implies I see her, while making somebody work does not imply making him (like one makes, for instance, a tool or cake). With deixar even the choice of verb changes in the English translation: Letting someone work as opposed to leaving someone. And it is this second reading of causatives we get with prototypical "predicative material", pp's, adjectives or adverbs:

    Fizeram-na famosa.

    (They made her famous.)

    Deixaram-no sozinho/em casa/sem comida.

    (They left him alone/at home/without food.)

    Thus, for the two causatives mentioned, if we choose the Oacc:fcl analysis for the make/let meaning, and the Co:pp/adj reading for the make/leave meaning, we now have a syntactic tool to distinguish between these two cases. However, while fazer only allows Co-predicatives, both deixar and the ACI sense-verbs permit Ao-predicatives (deixaram-no ., vi-o aqui).

    A third causative, the order-verb mandar, behaves even more like ACI-verbs: the pronoun substitution test (c-d) is positive, and Co-readings aren't even causative:

    (1) O rei mandou um soldado chamar a rainha. - O rei mandou o soldado.

    (2) O rei mandou o soldado sem armas. (Co)

    (3) O rei mandou o soldado à rainha. (Ao)

    The causative effect is stronger in (3) than in (2), since the soldier in (2)is without arms, he does not become without arms, whereas the soldier in (3) does end up with the queen. Therefore, um soldado in (1) could well be tagged as direct object (Oacc) followed by an icl object predicative. But which kind of object predicative, nominal (Co) as in (2), or adverbial (Ao) as in (3)? For ACI-verbs one can imagine neutrally sensing (for instance, watching) somebody who does something:

    Bate na mulher. Vejo isso. -> Vejo-o bater na mulher.

    This doesn't work with mandar:

    O soldado chama a rainha. O rei manda o soldado.

    Rather, as a cause-effect sequence the opposite is true: calling the queen is the purpose of sending the soldier, and a purpose subclause should be analysed consistently as adverbial no matter whether there is a causative matrix clause (2) or not.

    A third group of transobjective constructions are causatives that govern dative objects (e). Due to the case difference, it is even less satisfying to view the dative in these cases as subject of a direct object clause:

    Rather, one could use the same clause level functions (Odat/Opiv and Oacc) that appear in the concerning verbs' valency slots when filled with nominal material: permitir-proibir-aconselhar ac. a alg.:

    Transobjective constructions can occur even without the mediating accusative (or dative):

    If the "accusative" or "dative" is an np, not a pronoun, it can also appear to the right of the subclause predicator. This happens especially when the subclause main verb is ergative (i.e. governs a patient subject), since these verbs have a tendency to allow VO order:

    Mediator pronouns, on the other hand, can precede the matrix verb:

    In fact, pronoun fronting is a notational argument for not choosing an Oacc:icl analysis in transobjective constructions, since this would result in an accusative/dative subject to the left of a predicator whose subject it is not, a fact that in CG notation could be marked by a double dependency arrow:

    Also, the concerning syntactic tree would involve an (avoidable) disjunct constituent:

    Exercises:

    7.4-1. Do quarto, ouvi os outros saírem da casa.

    7.4-2. O rei mandou chamar os assaltantes.

    7.4-3. O rei mandou o delegado chamar os assaltantes.

    7.4-4. O rei mandou entrar os assaltantes.