Richards (2009) proposes that VP-Ellipsis relates pairs of sentences in the Eastern Algonquian language Maliseet-Passamaquoddy that contain morphologically related verbs, one of which lacks a stem component that appears in the other. Under his proposal, the polysynthetic verbs of the language are decomposed into abstract syntactic structures, and VP-Ellipsis is applicable within the word. I argue that his data are better explained without postulating VP-Ellipsis for Maliseet-Passamaquoddy. The verbs that he sees as derived via ellipsis are independently occurring lexical items. There are in fact no ellipsis gaps in the structures he considers, only null pronouns, and overt pronouns may always occur in the relevant positions. My conclusions are fully consistent with the Lexical Integrity Hypothesis of Bresnan and Mchombo (1995).
Richards (2009) proposes that the relationship between pairs of Maliseet-Passamaquoddy sentences like (1a) and (1b) reflects the application of VP-Ellipsis. Maliseet-Passamaquoddy is an Eastern Algonquian language of New Brunswick and Maine. Richards’s examples come from the Maliseet dialect of New Brunswick. Here I bring data to bear on his proposal from the Passamaquoddy dialect of Maine.1
N-ikuwoss ’t-apqote-htu-n khakon, kenuk nil nt-aluw-apqote-htu-n.
1-mother 3-open-TI-INAN door but I 1-unable-open-TI-INAN
‘My mother opened the door, but I couldn’t open it.’
N-ikuwoss ’t-apqote-htu-n khakon, kenuk nil nt-aluw-ehtu-n.
1-mother 3-open-TI-INAN door but I 1-unable-TI-INAN
‘My mother opened the door, but I couldn’t (open the door).’
Note that the verb of the second clause in (1a) contains an element -apqote-, repeated from the first clause, which Richards glosses as ‘open’. In (1b), we find a shorter verb in the second clause, one without the repeated -apqote-. I will use the term short-form verbs to refer to forms like this which seem to lack a piece of the stem that we might have expected to find repeated.
Richards proposes an abstract syntactic analysis of the sentences in (1) under which -apqoteis a verb.2 The absence of this element from the second clause in (1b) is then attributed to VPEllipsis. Richards sees this operation as deleting the noun khákon ‘door’ in the second clause in (1b) as well. Note also -ehtu-, glossed here as ‘TI’ for ‘transitive inanimate’. This suffix forms transitive verbs that belong to the class that select inanimate objects.3 (There is a corresponding class of transitive animate or TA stems that select animate objects. The distinction in question is a matter of grammatical animacy gender that does not always correspond to semantic animacy.) The suffix -ehtu- basically means ‘act on’ or ‘do’. We will see that it has a central role to play in the analysis of pairs of sentences like those in (1).
Richards’s analysis of these examples is potentially of considerable interest. The morphology of Maliseet-Passamaquoddy is polysynthetic, and the language is of the so-called nonconfigurational type (Hale 1983). That is to say, word order is highly flexible, null anaphora is routinely used, and discontinuous constituents are common. It has been suggested that clauses in polysynthetic or nonconfigurational languages lack VPs altogether or else that overt DP arguments of the verb are not located within the VP in such languages ( Jelinek 1984, Baker 1996). Demonstrating that Maliseet-Passamaquoddy has VP-Ellipsis would be significant, then, since it would show that the language has VPs. Moreover, if Richards is right that VP-Ellipsis in Maliseet-Passamaquoddy can delete verbs together with their objects, we would have proof that the VPs of this language contain the verb’s complements.
I argue, however, that Richards’s examples are better explained without postulating VP-Ellipsis for Maliseet-Passamaquoddy. First, I demonstrate that short-form verbs like ntaluwéhtun ‘I couldn’t do it’ in (1b) are independently occurring lexical items that have underspecified meanings and receive their full interpretations by pragmatic principles. This state of affairs renders VP-Ellipsis superfluous.
Second, I show that sentences like (1b) do not in fact contain gaps in the sense of ellipsis sites.4 The apparent gaps in this sentence and others like it are simply null pronouns. Overt pronouns are equally possible in all cases.
Moreover, there is a crucial fact about the syntax of verbs like ntaluwéhtun that Richards does not note: these verbs can take sentential complements. The null pronominal object of the verb in (1b) replaces such a complement. Its referent is a predicate abstracted from the first clause, here approximately λx [x open the door]. It is thus a sentential complement, not an abstract VP, that is ‘‘missing’’ in the second clause of (1b), and it is the presence of a pronominal object here that accounts for the interpretation that the verb receives: the reference to opening a door.
The conclusion that VP-Ellipsis is not involved in the derivation of sentences like (1b) has further implications. Richards’s approach to the analysis of Maliseet-Passamaquoddy draws no distinction between word structure and syntax, abandoning the Lexical Integrity Principle of Bresnan and Mchombo 1995 and other work in the Lexicalist tradition.5 This practice has become routine, almost standard, in theoretically oriented work on polysynthetic languages. Recent examples of this trend among studies of Algonquian languages include Branigan, Brittain, and Dyck’s (2005) analysis of verb stems in Innu-aimun, which postulates syntactic sources for verbs that include both phrasal and clausal structures, and Mathieu’s (2007, 2008) account of Ojibwe verbs, which likewise derives verb stems from highly articulated syntactic structures via phrasal movement. Compton and Pittman (2010) treat essentially all word formation in Eskimo languages as syntax.6
It is my contention that the view of word structure in polysynthetic languages that such work represents is on the wrong track. In the case of Richards’s analysis, it appears at first that a syntactic process, VP-Ellipsis, operates within the word. If this were indeed the case, we would have evidence that word structure is syntactic structure. But when we examine the facts in detail, we find that the issues that arise are better resolved through an account that does not postulate word-internal syntactic operations. Lexical Integrity is in fact respected in the best analysis of the phenomena in question. This conclusion casts doubt on the assumption that the structure of words is appropriately analyzed in syntactic terms.
2 Richards’s Proposal
Richards (2009) is not altogether explicit about the syntactic structures he has in mind, but the second clause in (1b), where VP-Ellipsis takes place, would evidently have the structure shown in (2). The circled VP would be the repeated structure targeted by ellipsis. This consists of -apqote- ‘open’, analyzed as a verb, plus khákon ‘door’, taken to be its object. Other components of the verb stem are given analyses as syntactic constituents as well: the transitivizer -ehtu- is analyzed as little v, while the stem component aluw- ‘try and fail, be unable’ is analyzed as a verb heading a VP complement of v. This verb undergoes raising to v, which attaches it to the left of -ehtu and the suffixes it bears. The resulting complex then raises to join the prefixal inflection in T, giving the surface order of morphemes.
Richards (2009:246–247) argues that support for his hypothesis comes from the fact that the complement of V within a VP that undergoes ellipsis may trigger object agreement. This effect can be seen most clearly where this object is plural, as it is in (3), where the suffix -ol marks the short-form verb of the second clause as agreeing with an inanimate plural object.
N-ikuwoss ’t-apqote-htu-n-ol khakon-ol, kenuk nil nt-aluw-ehtu-n-ol.
1-mother 3-open-TI-INAN-INAN.PL door-INAN.PL but I 1-unable-TI-INAN-INAN.PL
‘My mother opened the doors, but I couldn’t (open the doors).’
(Maliseet; Richards 2009:247, (10b))
3 An Alternative Proposal
The alternative to Richards’s analysis that I wish to propose takes (1b) and (3) to be instances of two distinct constructions, neither of which contains a gap. Both involve verbs in -ehtu- ‘act on, do’, whose stems I take to be formed in the lexicon. The stems in question have general or underspecified meanings, receiving specific interpretations in context by pragmatic principles.7 Verbs in -ehtu- are transitive and select inanimate objects, which may be DPs or clauses.
The first construction to consider, that seen in (1b), arises when the complement of a verb in -ehtu- is a null pronoun with the semantic force of a clause. Consider in this connection the examples in (4).
‘I was unable to stop smoking.’
‘The boy painted some rocks, and I did it (= painted rocks), too.’8
There are two verbs in -ehtu- here. Both are transitive forms taking inanimate objects, as shown by the fact that both bear the inanimate object agreement suffix -n. In (4a), we have a form of aluw-ehtu- ‘be unable to do’, the same verb as in (1b). The object of the verb in this case is the clause ‘I stop smoking’. The corresponding object in (4b) is pronominal.9
If the optional pronominal object is omitted in (4b), then this sentence has an interpretation that suggests ellipsis, since nkiséhtun ‘I did it’ is understood in its context here as ‘I painted rocks’. But there is no ellipsis gap in (4b), only an optionally null object. The reading that the verb receives is a function of the force of this pronominal object, which refers to the activity of rock painting named in the first clause. Note, too, that the semantic force of this object is parallel to that of the sentential complement of the verb in (4a).
The second construction that gives the appearance of ellipsis arises because verbs in -ehtumay take a single DP complement in syntax, but also include an implicit propositional complement in their semantics. For example, the verb aluw-ehtu- has the reading ‘do (something) to an inanimate’ in (5), where the ‘something’ that is done is determined from context to be opening doors.
Note that there is no ellipsis gap in (5): the DP object of the short-form verb in the second clause may be overtly expressed. But when this object is left unexpressed, simply through null anaphora, then the example presents the appearance of an ellipsis structure and the verb looks like it has agreed with the object in an ellipsis gap.
This is the thrust of the proposal for which I argue in the sections that follow. The apparent ellipsis structures that Richards identifies in Maliseet do not in fact contain gaps. The features that make them appear to involve ellipsis follow from the syntactic, semantic, and pragmatic properties of verbs in -ehtu- ‘act on, do’.
4 Richards’s Central Arguments
Richards offers two central arguments in favor of his analysis, one based on the interpretation of adverbs, the other on the readings assigned to direct objects.
First, he notes parallels with respect to the way adverbs are interpreted between the structures in Maliseet-Passamaquoddy that he takes to be elliptical and VP-Ellipsis structures in English. The examples in (6) are typical.
N-ikuwoss menakaciw ’t-apqote-htu-n khakon, kenuk nil nt-aluw-apqote-htu-n.
1-mother quietly 3-open-TI-INAN door but I 1-unable-open-TI-INAN
‘My mother opened the door quietly, but I couldn’t open it.’
N-ikuwoss menakaciw ’t-apqote-htu-n khakon, kenuk nil nt-aluw-ehtu-n.
1-mother quietly 3-open-TI-INAN door but I 1-unable-TI-INAN
‘My mother opened the door quietly, but I couldn’t (open the door quietly).’
When the full verb ntaluwapqotéhtun ‘I couldn’t open it’ appears in the second clause, as it does in (6a), the adverb menakacìw ‘quietly’ is taken to apply only to the first clause: ‘My mother opened the door quietly, but I couldn’t open the door’. There is no implication that I was unable to be quiet. But when the short-form verb ntaluwéhtun, literally ‘I couldn’t do it’, appears in the second clause, this clause has the reading ‘I couldn’t open the door quietly’. This situation parallels what we find with VP-Ellipsis in English, as in the glosses in (6), and Richards suggests that a parallel analysis is called for: namely, one in which there is an underlying VP in the second clause that matches the VP in the first clause in containing the adverb.
Second, Richards notes that the interpretation of the sentences in (7) parallels that of their English counterparts. In (7a), the object of the verb is a null pronoun, which is interpreted as referring to the previously mentioned rock. In (7b), Richards sees the stem sunhom- ‘paint’ and the verbal object ponápsq ‘rock’ as subject to VP-Ellipsis. Since the object is deleted rather than pronominalized, he reasons, it can be understood to refer to a different rock from the one mentioned in the first conjunct.
Skinuhsis ’-kisi=sunhom-on ponapsq;
boy 3-PERF=paint-INAN rock
‘The boy painted a rock . . . ’
Let us consider this second set of data first. In point of fact, I find that the reaction of my Passamaquoddy consultants to examples like those in (7) varies. Two speakers consulted agree with the judgments that Richards reports and take (7b) to mean that I have painted a different rock from the boy mentioned in the first clause. A third finds that (7b) means that I have painted the same rock as the boy. Both readings can in fact be accommodated under the alternative to Richards’s proposal that I have outlined above.
I have suggested that there are two constructions that give rise to the appearance of ellipsis in sentences like those we have been considering. In the first, a lexically created short-form verb in the second of two clauses takes a pronominal object, possibly null, that refers to the situation specified in the first clause. In (7b), this verb is nkiséhtun, a transitive verb meaning not ‘I did’, but ‘I did it’. In other words, the sentence we are dealing with is approximately ‘The boy painted a rock, and I did it, too’. The reference of ‘it’ is roughly ‘paint a rock’; no particular rock is specified; and the Maliseet or Passamaquoddy sentence may be understood, like its English counterpart, to indicate that I painted a different rock from the boy in question.10
In the second construction, nkiséhtun takes a DP with concrete reference as its object and is understood to mean ‘I did something to it’. In this case, the identity of the ‘something’ done is determined from context to be rock painting, and ‘it’ is taken to refer to the particular rock already mentioned, just as in (7a). So the resulting reading is the same as that for (7a): the speaker has painted the same rock as the boy. This type of interpretation is disfavored by one of my consultants for the particular verb kis-ehtu- ‘acted on, did (something) to’. This speaker insists that the more complex form kisi=l-ehtu- (PERF=thus-act.on-) ‘acted on thus’ is required to achieve this kind of reading.11 Thus, the only interpretation that this speaker allows for (7b) is one in which I have painted a different rock from the boy, in agreement with Richards’s consultants. His brother shares his judgments on the ‘painting’ examples.
Let us return, then, to (6), where an adverb in the first of two clauses is taken to be applicable to the second because the latter includes a short-form verb. This effect follows again from the fact that a verb in -ehtu- may take a pronominal object that represents a predicate identified by reference to the first clause in the sentence. Richards’s gloss of (6b) with an English VP-Ellipsis structure is misleading here. A better gloss would be ‘My mother opened the door quietly, but I couldn’t do it’. As before, the referent of the pronoun in the second clause (the null object of the verb ntaluwéhtun ‘I was unable to do it’) is obtained by abstracting over the first clause, giving roughly λx [x open the door quietly]. The translation of the adverb ‘quietly’ is then carried over from the interpretation of the first clause in (6b) to that of the second via this interpretation of the object in the latter. We do not need to appeal to ellipsis to achieve this end.
5 Arguments for the Alternative
The analysis of apparent ellipsis structures in Maliseet-Passamaquoddy that I have been advocating makes two central claims. First, short-form verbs are not derived by ellipsis, but are simply basic lexical items with underspecified meanings whose interpretation becomes specified in context by pragmatic principles. Second, there are no ellipsis gaps in the structures we are concerned with here, only null pronouns. The apparent gap in (1b), the example with which we began, is a null pronominal that stands in for a clause.
To see that only a pragmatic analysis can account for the interpretation of short-form verbs in the general case, consider (8).
‘She stopped by to visit her friend, but when she tried the door there she couldn’t open it.’
(Passamaquoddy; Francis and Leavitt 2008:63, accent marking supplied)
There are two verbs in -ehtu- in this sentence: ‘try to do’ and ‘be unable to do’. The first is interpreted in context as ‘try to open’, in reference to a door, since it is used here to describe what someone does upon stopping at a friend’s house. Note that ‘be unable to do’ in the next clause is precisely the verb aluw-ehtu- that Richards takes to be the output of VP-Ellipsis in example (1b). And indeed the sense of this verb in (8) is once again ‘be unable to open the door’.
But how does this sense arise in this case? The idea of opening a door is introduced in the preceding clause only indirectly, by saying ‘try the door’. There is no verb in the linguistic context that includes a morpheme meaning ‘open’. Thus, there is no question of VP-Ellipsis here.12 The interpretation ‘she was unable to open the door’ can only be pragmatically determined. But if pragmatics can account for this interpretation of the verb ‘unable to do’ in a context like that in (8), then there is no need to appeal to VP-Ellipsis to account for the interpretation of this verb in (1b), either. I conclude that the appearance of VP-Ellipsis in Maliseet-Passamaquoddy sentences like (1b) is simply a result of the interpretation in context of verbs with an underspecified meaning.
The nature of the empty positions that occur in apparent ellipsis structures is revealed by examples like those in (9). In (9a), we have a Passamaquoddy analogue of Richards’s Maliseet example (1b). The short-form verb ntaluwéhtun ‘I was unable to do it’ is interpreted as ‘I was unable to open the door’, giving the impression of VP-Ellipsis. In (9b), however, we see what is actually ‘‘missing’’ in (9a): the entire clause ‘I open that door’. The object of the verb in (9a) is a null pronoun with the semantics of a clause, referring to the situation in the first clause and giving this sentence the reading whereby ‘unable to do it’ is interpreted as ‘unable to open that door’.
N-íkuwoss ’t-apqoté-htu-n ’-kahákon-um . . .
1-mother 3-open-TI-INAN 3-door-POSS
‘My mother opened her door . . . ’
. . . qénoq nìl nt-aluw-éhtu-n.
but I 1-unable-TI-INAN
‘ . . . but I couldn’t do it (= open the door).’
. . . qénoq nìl nt-aluw-éhtu-n [CP nt-apqoté-htu-n nìt khákon].
but I 1-unable-TI-INAN 1-open-TI-INAN that door
‘ . . . but I couldn’t open that door.’
As we observed in section 3, the pronominal object in the construction in (9a) need not be null. In (10), the pronominal object of the short-form verb is optionally realized as an inanimate demonstrative—providing direct evidence that there is no ellipsis gap in the structure illustrated here.
On my analysis, the pronominal object of ntaluwéhtun ‘I couldn’t do it’ in (10) has the same status as the clausal complement of this verb in (9b). Richards (2009:251–253) takes a different approach to sentences like (10), however. He considers only variants with null pronominal objects, giving these an analysis in terms of VP-Pronominalization. That is, he takes the antecedent of the null object in the second clause of a sentence like (10) to be a VP in an abstract syntactic representation for the first clause in a structure like that shown in (2).
Here ntaluwéhtun (nìt) has the reading ‘I can’t do that’, much as before, but ‘that’ is understood to refer to the conjunction of two activities, singing and dancing. How is this reading established?
When two verbal expressions are conjoined in Maliseet-Passamaquoddy in a nonembedded context like that in (11), the verbs are inflected differently: the first receives indicative inflection, in which the third person is marked by a suffix (underlying /-w/), and the second receives subordinative inflection (signaled by a suffix -n), in which the third person is marked by a prefix (underlying /w(t)-/).13 For Richards, such inflectional differences are reflected in T. Thus, the two verbal expressions in (11) must be associated with distinct T nodes. They can accordingly be conjoined no lower than the T′ level. If we follow Richards’s lead and take the stem-final components of the verbs here to be instances of little v, then the representation of the first clause in (11) will be as in (12).
Note that ‘sing’ and ‘dance’ do not form a VP together in (12). Indeed, neither of these elements is dominated by a VP node. Even if we analyze ‘sing’ and ‘dance’ as verbs within VP rather than as instances of v, these items will not form a constituent together, since they are dominated by different T′ nodes. Thus, there is no analysis of (11) under which ‘sing and dance’ constitutes a VP that can stand as the antecedent of the pronoun nìt ‘that’ (or its null counterpart) in the second clause of this sentence on a VP-Pronominalization account.14 I conclude that such an account is untenable. The alternative analysis offered here postulates pronouns in the structures in question with pragmatically determined reference. The proposition ‘I sing and dance’ is certainly pragmatically accessible as the antecedent of the object pronoun in (11), whether this form is overt or null.
6 The Construction with a DP Object
As noted in section 2, Richards suggests that a verbal object that he takes to be deleted by VP-Ellipsis can control object agreement. Example (13), repeated from (3), is a case in point. Here the suffix -ol marks the verb in the second clause as agreeing with a covert inanimate plural DP.
I have argued that such examples arise without ellipsis when a lexically derived short-form verb takes an ordinary DP object that happens to be pronominal, hence null.15 Evidence for this approach comes from Passamaquoddy sentences like (14), in which the object of the short-form verb is a lexical DP, here discontinuously expressed: the structure of (13) is presumably parallel in relevant respects to that of (14).
Richards does not discuss Maliseet examples like (14), but several Maliseet speakers with whom I have worked report that analogues of this sentence are fully acceptable in their speech. The speakers consulted in this connection include two from the St. Mary’s Reserve at Fredericton, New Brunswick, one from Kingsclear, and one from Tobique, where Richards carried out his research. Thus, their speech is representative of the diversity found within the contemporary Maliseet speech community.
On the analysis that I have presented, there is nothing to prevent a sequence of clauses from arising in which a full verb in the first clause is matched by a short-form verb in a second clause, but the two verbs have distinct DP objects. It is therefore a puzzle for my analysis that Richards’s Maliseet consultants reject examples like (15), which displays just this configuration: the verb ‘closed’ in the first clause takes ‘door’ as its object, the short-form verb ‘unable to do (something = closing) to it’ in the second clause takes ‘window’ as its object, and the sentence is found wanting.
In fact, however, my Passamaquoddy consultants find the Passamaquoddy analogue of (15) to be entirely acceptable. Moreover, all of the members of my group of Maliseet speakers unhesitatingly judged (15) to be entirely natural. Thus, the reluctance of Richards’s consultants to accept the construction represented by this sentence does not seem to be general among speakers of the Maliseet dialect.
I note finally that the account of apparent ellipsis structures that I have advanced would lead us to expect to find analogues of examples like (13) and (14) above (where a short-form verb takes a concrete object) in which the object is grammatically animate. The short-form verbs in such sentences would have stems formed with the suffix -ehl- ‘act on (thus), do (so) to’, which derives TA verbs, transitive verbs that select animate objects. In Passamaquoddy, we find examples of exactly the predicted form, as illustrated in (16). The verb here is kis-ehl- ‘did (something) to an animate’; its object is a null pronoun referring to the grammatically animate noun amsqocéhkan ‘doll’.
Richards (2009:254) again reports a different situation for Maliseet, indicating that his consultants reject (16). He sees this fact as evidence for an animacy restriction on the process of VPEllipsis. But when I consulted three members of my Maliseet panel, they failed to confirm the judgments that Richards had obtained. Like my Passamaquoddy consultants, they agreed in finding (16) fully natural. Once again, then, it seems unlikely that there is any significant difference here between the Maliseet and Passamaquoddy dialects. In any case, the evidence for a general constraint of some kind appears weak.
I have investigated two constructions that give rise to the appearance of ellipsis in Maliseet- Passamaquoddy. Both involve verbs formed with -ehtu- ‘act on, do’ that have underspecified meanings. I have called these short-form verbs because they occur in contexts where they seem to be ‘‘missing’’ components that occur in a morphologically related verb in a preceding clause. It is this morphological relationship that has suggested to Richards (2009) that VP-Ellipsis is involved in the derivation of such forms. But in fact, the same verbs occur in contexts where no relevantly related form precedes them. In such cases, their meanings are clearly pragmatically determined. This fact renders an ellipsis account of their derivation in seeming ellipsis contexts superfluous.
But we can say more than this. VP-Ellipsis leaves a gap. But as I have demonstrated repeatedly, wherever Richards’s analysis would postulate a gap, it is possible to have an overt DP. This DP may be a pronoun alternating with a clause, in the case of the first construction that I have identified, or it may be a full DP, in the case of verbs that take a concrete object. Since there are no VP-Ellipsis gaps in the constructions under consideration, only null pronouns, there is nothing to suggest that VP-Ellipsis is involved in their derivation.
I conclude that the case for VP-Ellipsis in Maliseet-Passamaquoddy fails. In the end, then, there is no motivation from ellipsis phenomena for the syntactic decomposition of Maliseet- Passamaquoddy verbs. On the contrary, the best analysis of the constructions to which Richards has called attention is fully compatible with the Lexical Integrity Principle of Bresnan and Mchombo 1995.
I owe special thanks to the two Passamaquoddy elders who worked most closely with me in connection with the present project, Estelle Neptune and Wayne Newell. Passamaquoddy consultant Phillip Newell made important contributions as well. Maliseet speakers Victor Atwin, Sharon Nash, Darryl Nicholas, and Alice Paul provided assistance with material in their language. Paul Kroeber, Kevin Rottet, and Rex Sprouse commented helpfully on drafts of this article, and two LI reviewers made suggestions that led to substantial improvements.
1 Examples are given in a modified version of a widely used standard orthography: o represents /ə/, u is /o/, c is /č/, and q is /kw/. Phonemic /h/ before a consonant at the beginning of a word is written as an apostrophe. The acute accent indicates high pitch and the grave accent low pitch for the distinctively accented vowel in a word. Accent is not marked, however, in examples taken from Richards 2009. The double hyphen = is used to join an enclitic particle to its host and to connect a preverb to a following verb.
The following abbreviations are used in glosses: 1 = first person, 3 = third person, AN = animate, DIR = direct, EMPH=emphatic, INAN=inanimate, OBV=obviative, PERF=perfect, PL=plural, POSS=possessed, PRET=preterite, SG = singular, SUB = subordinative, TA = transitive animate, TI = transitive inanimate.
2 This item -apqote- is in fact morphologically complex, consisting of apq- ‘open’ and -ote- ‘covering’: compare ’t-apq-éhtu-n (3-open-TI-INAN) ‘he unties, unfastens it’, ’kopp-ote-hté-hm-on ((3)-close-covering-strike-TI-INAN) ‘he boards it (house) up’. Since the details of morphology are not our focus here, however, I will generally follow Richards’s glosses.
3 This transitivizer -ehtu- consists of a stem-forming suffix -eht- plus a so-called thematic element -u-, the latter occurring with a number of transitivizing suffixes; compare macé-pt-u-n ‘he carries it away, he leaves with it, he takes it home’, with mace- ‘start’ and -pt- ‘carry’.
4 I use the term gaps to refer to positions from which material has been removed by ellipsis, distinguishing such positions from others that are empty through the use of null pronouns.
5Bresnan and Mchombo (1995:181) state their view of lexical integrity in the following terms:
[T]he lexical integrity principle . . . states that words are built out of different structural elements and by different principles of composition than syntactic phrases. Specifically, the morphological constituents of words are lexical and sublexical categories—stems and affixes—while the syntactic constituents of phrases have words as the minimal, unanalyzable units; and syntactic ordering principles do not apply to morphemic structures.
7 Only certain stems formed with -ehtu- have underspecified meanings. In most cases, the meaning of stems in -ehtu- is specified by the contributions of other stem components: ’t-apuckol-éhtu-n ‘he turns it upside down’ (apuckol- ‘upside down’), ’t-ewep-éhtu-n ‘he lifts it, hoists it’ (ewep- ‘upward’), ’t-uwapol-atok-éhtu-n ‘he wires it or strings it incorrectly, he gets it (story) mixed up’ (uwapol- ‘incorrect’, -atok- ‘stringlike object’).
8 For the consultant from whom this example was obtained, the sentence can only mean that I painted the same rocks as the boy in question. See footnote 10.
9 There are no inanimate personal pronouns in Maliseet-Passamaquoddy. Thus, the only type of inanimate pronoun that may appear in a context like that of the object in the second clause in (4b) is a demonstrative.
10 The speaker for whom (4b) means that I painted the same rocks as the boy is the speaker for whom (7b) means that I painted the same rock as the boy. For this speaker, it appears that the ‘it’ of ‘I did it, too’ in the second clause of these examples is interpreted not as ‘paint a rock’ or ‘paint some rocks’ but as ‘paint the rock(s)’, with specific reference to the rocks mentioned in the first clause. It is not clear why there should be such variation among speakers.
11 The effect of adding the element l- ‘thus’ to the verb here is to introduce an overt morpheme to represent the propositional argument that is only implicit in kis-ehtu- ‘did (something) to an inanimate’, where ‘something’ refers to the event named in the preceding clause.
12Larson, Den Dikken, and Ludlow (1997) argue that apparent NP objects of intensional transitive verbs like ‘want’ and ‘need’ actually appear in covert clausal complements, so that ‘want a beer’ and ‘want to have a beer’ are parallel in structure. A reviewer wonders whether ‘try the door’ in (8) might be similar, with a structure like [VPtry [VP V [NPthe door]]], where V is an abstract verb with the meaning ‘open’, determined by the context. There could then be an antecedent for VP-Ellipsis in the following clause. But in fact the Maliseet-Passamaquoddy verb that renders ‘try’ in ’t-oqec-éhtun khákon ‘he tries the door’ (lit., ‘he tries to do the door’) already includes a concrete morpheme corresponding to this V, namely, -ehtu- ‘act on, do’. An entirely parallel verb form renders (for example) ‘try to make’ in ’t-oqec-ìhtu-n posonùt ‘he tries to make the basket’ (with -ihtu- ‘make’), which is expressed in English with a complement clause. Thus, there is no reason to postulate an abstract verb in the Maliseet-Passamaquoddy structure for ‘try the door’. Note, however, that the interpretation of -ehtu- ‘do’ as ‘open’ in (8) must be pragmatically determined, as indicated in the text.
13 Compare kséhe naka (w)t-ópi-n ‘he came in and sat down’.
14 A reviewer suggests that we might postulate a structure here of the form [VPlike [VPsing and dance]], with VPPronominalization targeting the lower VP. Then ‘like’ could take the form of an affix that is distributed over ‘sing’ and ‘dance’, with these verbs functioning as parallel morphological hosts, somewhat the way -er distributes over two hosts in English picker-upper. Such an analysis is excluded, however, by the fact noted in the text that the two verbs in question are associated with distinct T nodes, ruling out conjunction here at the VP level or below.
15Richards (2009:247) also notes that an object that he takes to be deleted by VP-Ellipsis can be relativized. On the alternative analysis presented here, this fact is expected. Such examples are again instances of the construction in which a short-form verb takes a concrete object, which may be relativized like any other DP.