Embedding, covert movement, and intervention in Kathmandu Newari

In this paper, we explore the syntax of wh-dependencies in Newari (Sino-Tibetan). We examine the patterns of intervention and island effects in wh-in-situ configurations, and we find that sensitivity to these constraints often co-occur. We thus argue that Newari permits wh-operators to either covertly move to fix their scope, or may take scope in-situ via focus alternative composition analysis. Additionally, we argue that clausal complements to verbs (“verbal argument CPs”) may be islands for covert movement in this language.

1.1.THEORETICAL BACKGROUND.In some languages, wh-operators in simple interrogatives overtly move, and in others, the wh-operator stays in-situ.In languages like English, the wh-operator moves from the base position to the Spec,CP position to take sentential scope, shown in (1).In languages like Newari, as in (2), the wh-operator may take sentential scope in-situ.
(1) What i did Ram eat t i ? (English) (2) Rām-na Ram-ERG chu what na-la?eat-PST 'What did Ram eat?' (Newari) Broadly, there are two main analyses of how in-situ wh-operators take wide scope without overt movement.One approach posits covert movement of the wh-operator (Huang 1982, Soh 2005, Yang 2012).This approach suggests that the wh-operator stays in-situ on the surface syntax as in (3a), but moves to the Spec,CP position at LF, as in (3b).
(3) a. Covert movement (CM) analyses prima facie predict that wh-in-situ configurations should exhibit the same properties as overt movement, e.g., island sensitivity.In fact, island sensitivity is observed in wh-in-situ configurations.For example, in Mandarin Chinese, in-situ wh-operators display sensitivity to the Complex NP Constraint (Huang 1982, Bayer 2006, Cheng 2009).As shown in (4), the adverbial wh-phrase weishenme 'why' cannot covertly move out the DP, and thus fails to take sentential scope.(Mandarin, Huang 1982) Another account of wh-in-situ configurations is the focus alternatives composition analysis (FA).On this approach, wh-phrases are focus elements, and are interpreted by computing the focus semantic alternatives of the sentence (Beck 2006, Hamblin 1973, Karttunen 1977).On this account, wh-phrases have no ordinary semantic value, but instead "shift" the semantic value via (g) into a focus tier (g, h), until recombining with a morpheme Q in the root clause at LF, as in (6).
Crucially, this analysis suggests that the interpretation of wh-operators does not involve any movement.
[ TP Ram ate what] g = undefined One advantage of the FA approach is that it provides a natural explanation for intervention effects, which are observed in many wh-in-situ configurations (Beck 2006).Intervention effects are the unacceptability that arises when an in-situ wh-operator appears within the scope of a focus-sensitive operator, demonstrated in the sentences in ( 7) and ( 8).These sentences are ungrammatical, because the in-situ wh-operator appears within the scope of the intervener 'only'.According to Beck (2006), intervention effects arise because the focus-sensitive operator cannot combine with the constituent containing the in-situ wh-operator, since the ordinary semantic value of this constituent is undefined, demonstrated in (9).( 7 [buy what 1 ] g = undefined b. [buy what 1 ] g,h = λw.λxx bought h(1) in w c. * only John g ( [buy what 1 ] g ) 1.2.THE PUZZLE IN NEWARI.Generally, the CM and FA approaches are understood as alternative analyses for analyzing the scope of in-situ wh-operators.If in some configuration, an in-situ wh-operator fails to take sentential scope, this may be understood as arising from (island) constraints on covert movement, or by intervention effects constraining focus composition.However, our Newari data show a different pattern: neither island effects or intervention effects occur in matrix clauses, whereas both occur in certain embedded clauses.We will lay out the puzzle in this section and explain this matter in detail in Section 2 and 3, and suggest an account by combining the two approaches in Section 4.
First, intervention effects are not observed for in-situ wh-operators in Newari matrix clauses.In contrast to the Hindi and Korean examples in ( 7) and ( 8), Newari wh-operators may take sentential scope over a focus-sensitive operator in matrix clauses, demonstrated in (10).We suggest that the wh-operator takes scope through CM in this case, given the lack of intervention effect.
Although we do not observe intervention effects in matrix clauses, we do find them in embedded clauses.In (11a), a wh-operator may take either wide or narrow scope in an embedded clause.However, as shown in (11b) the wide-scope reading is blocked when the wh-operator is in the scope of another focus operator.
Sita-ERG Given that there is no clear focus-sensitive operator in this sentence, we argue that this is an instance of an island violation for covert movement.
The pattern that we seek to explain is summarized in Table 1.On most accounts of wh-in-situ configurations, we do not expect island sensitivity and intervention to co-pattern in this way.In this paper, we investigate the following questions: 1) Why does the distribution of wh-phrases determine the constraints that they are sensitive to? 2) Why do we not observe traditional island constraints on covert movement in Newari? 3) Why does Newari exhibit these patterns, but not previously studied languages (Mandarin Chinese, Japanese, Hindi, . . .)?We also propose that Newari only allows the FA strategy for in-situ wh-operators in verbal-argument CPs, as in (15).We will show some evidence to support this analysis in the next two sections, and then discuss a complex case of a non-verbal-argument CP embedding in a verbal-argument CP, where both FA and CM may be deployed.( 15) 2. Non Verbal-Argument CPs.In this section, we examine the behavior of in-situ wh-operators in matrix clauses, and embedded clauses that are not arguments to verbs.We demonstrate that both FA and CM strategies may be used in these contexts.
As described above, Newari is typically a wh-in-situ language, i.e., wh-operators appear in their canonical position and take sentential scope, as demonstrated in (16a) and (16b).Wh-operators are homophonous with indefinite pronouns, although they may be assigned a higher tone when interpreted as a wh-operator.Additionally, wh-operators may scramble to the beginning of the clause, as demonstrated in (17a).This is improved if the wh-operator is D-linked, shown in (17b).( 16 As mentioned in the previous section, we propose that wh-operators in matrix clauses may either use FA or CM strategies.First, we do not observe intervention effects, either for argument wh-operators, as demonstrated in (10), or for adjunct wh-operators, demonstrated in (18).In these sentences, the wh-operators are c-commanded by the focus operator caka 'only'.If the wh-operator was interpreted with sentential scope through an FA strategy, then these sentences would be predicted to be ungrammatical.Therefore, we propose that the CM strategy is necessary for these configurations.
(18) Rām-na-caka Ram-ERG-only chae why am .mango.ABS na-u?eat-PST 'Why did only Ram eat a mango?' Next, we turn our attention to adjunct clauses adjoined in matrix clauses.We investigate island sensitivity for in-situ wh-operators.Huang (1982) argues that in-situ adjunct wh-operators are typically ungrammatical in traditional island configurations -relative clauses, adjunct clauses, comparative clauses, etc.As the example shows in ( 19), the wh-operator weishenme ('why') is sensitive to the complex NP constraints.However, there is an asymmetry between wh-adjuncts and wh-arguments in Mandarin Chinese.Wh-arguments are not sensitive to island constructions, demonstrated in ( 20 In Newari, we find no sensitivity to traditional islands for in-situ wh-operators.We find that both argument wh-operators, as in ( 21) and ( 22), and adjunct wh-operators, as in ( 23), may take wide scope from relative clauses, comparative clauses, and adjunct clauses.To account for these facts, we could suggest that the wh-operator takes sentential scope through an FA strategy.If so, we predict no island sensitivity, as observed above.However, we then predict that we should find intervention effects for these configurations.Surprisingly, as we show below, this prediction is not borne out.Thus, the wh-operator must take scope through a CM strategy, and we conclude that relative clauses, comparative clauses, and adjunct clauses are not islands for covert movement of wh-operators.Importantly, we find no intervention effects for wh-operators in these positions.In (24), we find that suna 'who.ERG' takes wide scope, even though it is interpreted in the scope of caka 'only'.This implies that suna does not take sentential scope through an FA strategy.Instead, we infer that it fixes its scope through the CM strategy.Similar findings are demonstrated in (25).( 24 In this section, we argued that wh-operators must covertly move in Newari, given that there are no intervention effects observed for matrix wh-in-situ configurations.Furthermore, we found no evidence of island effects for covert movement, in many traditional island configurations, specifically, relative clauses, comparative clauses, and adjunct clauses. 3. Verbal-Argument CPs.We now turn our attention to the verbal-argument CP structures.In this section, we examine whether island effects or intervention effects are observed in this kind of structure, and whether argument wh-operators or adjunct wh-operators show different profiles.We also find that a sentential scope interpretation is available if the wh-operator overtly scrambles above the focus-operator, as in (28).This supports our analysis, since overt movement has been independently shown to ameliorate intervention effects (Beck 2006 Furthermore, verbal-argument CPs can be fronted.If so, the wh-operator can take sentential scope, as in (29).Moving a verbal-argument CP over an intervener can ameliorate an intervention effect.In (30a), we find obligatory embedded scope for chu 'what', which we attribute to an intervention effect induced by caka 'only'.However, moving the CP to the front of the clause permits both sentential and embedded scope, as shown in (30b).( 29 We find the same result for adjunct wh-operators, shown in (31a).An intervener forces the wh-operator chae 'why' to take embedded scope.However, moving the verbal-argument CP to the front of the clause permits sentential scope.

INTERVENTION EFFECTS
(31) a.To summarize, we demonstrate that we consistently get intervention effects for wh-operators embedded in verbal-argument CPs, as sketched in (34), unlike matrix clauses or in other embedded CPs.On a CM analysis, these findings are surprising, since covert movement should ameliorate intervention effects in the LF representation.In this paper, we argue that, in general, Newari employs both FA and CM strategies for in-situ wh-operators, and verbal-argument CPs must be islands for covert movement in Newari.✗ However, there are some apparent counterexamples to our proposal.On our analysis, wh-operators embedded in verbal-argument CPs must rely on an FA strategy to be interpreted with sentential scope.This is because covert movement out of verbal-argument CPs is blocked.If so, then we predict that there should be no sensitivity to traditional island configurations, since only (covert) movement is sensitive to island constraints, by hypothesis, and because we have argued that traditional island constraints do not apply to covert movement in Newari.However, as demonstrated in ( 35) and ( 36), it is ungrammatical to have a wh-operator in a relative clause embedded in a verbal-argument CP.This might be interpreted as an island violation, given that relative clauses and adjunct clauses are typically islands.If so, then this appears to contradict our proposal that relative clauses and adjunct clauses are not islands in Newari, and that only the FA strategy is employed for wh-operators embedded in verbal-argument CPs, as sketched in (37).Since Newari employs both FA and CM strategies, it is unclear why they may not be applied on after the other (CM after FA) for the wh-operator to take sentential scope, as suggested for English by (Kotek & Hackl 2013), sketched in (38).Such a derivation appears to be unavailable for Newari wh-in-situ.
(38) Model of English multi-wh-questions: 4. Towards to an Account: A combination of Covert Movement and Focus Alternatives in Newari.In this paper, we proposed that, in Newari, (1) the mechanism for fixing wh-scope depends on the configuration that the wh-operator surfaces in, and (2) verbal-argument CPs are islands for covert movement, but not relative clauses, comparative clauses, or adjunct clauses.
Given the absence of intervention effects in matrix clauses, we proposed that wh-operators covertly move in Newari.However, in principle, FA may also be available.Thus, the sentence in (39) may be syntactically ambiguous as in (40a) and (40b).We propose that Universal Grammar permits wh-operators to take scope either by either the CM or FA strategies.b.Focus alternatives composition LF: If so, we reasoned that the presence of intervention effects for wh-operators in verbal-argument CPs must arise because the CM strategy is blocked, i.e., verbal-arguments are islands for covert movement.At this point, we do not have a clear explanation for why this should be the case in Newari, but not in many other well-studied languages, such as Mandarin, Japanese, or Hindi.One possibility is that these constraints follow from the syntax of evidentiality in Newari.Newari uses a "conjunct/disjunct" agreement system, in which the verb agreement is determined by whether the subject is coreferential with the perspective-holder (Zu 2015, Coppock & Wechsler 2016).Following Zu (2015), we assume verbal agreement is controlled by a null pronoun in Spec,CP that is coindexed with the perspective holder.Conjunct agreement occurs when the grammatical subject and this pronoun are co-indexed, and disjunct agreement occurs when they are contra-indexed: (41) a. Wo-m .Newari (Hargreaves 1991) If this account is on the right track, then it may be that this perspective-taker pronoun in Spec,CP may block the Spec,CP escape hatch.In other words, the availability of conjunct/disjunct agreement in Newari may induce a subjacency violation for covert movement from the verbal-argument CP.This appears to contradict the findings from Huang (1982), however, and thus needs to be examined in more detail in future research.
Relatedly, we are forced to say that traditional islands (CNPC, relative clauses, adjunct clauses) are not islands for covert movement in Newari, although verbal-argument CPs are.However, overt movement (e.g., relativization) appears to pattern in the expected waysextraction from traditional islands is unacceptable, and extraction from embedded clauses, as showing in (42a) and (42b), while (43) seems to be better than the former ones.Finally, this approach over-generates.Recall that wh-operators in an adjunct clause embedded in a verbal-argument CP cannot take sentential scope, which we diagnosed as an island effect.However, if a focus alternative analysis is always available in principle, this sentence is predicted to be grammatical.Thus, we need some way of blocking focus alternatives in these contexts, even though it seems necessary in other contexts, as (44) shows: (44) * ākās-ām .
Akash-ERG dhā-u?say-PST 'For which person x, Akash said that Ram met the teacher y that x met y.' Despite these shortcomings, the evidence in Newari suggests that covert movement and focus alternatives are both available, each with their own locality constraints.We argue that the findings that we've described follow as a conspiracy from these constraints, plus independently-motivated grammatical properties (i.e., conjunct/disjunct agreement).

Conclusion
In this paper, we demonstrated that Newari non-argument CPs (matrix clauses and adjunct clauses) do not show intervention effects or island effects, while intervention effects are observed in the verbal-argument CPs, and island effects appear when adjunct clauses are embedded inside of the verbal-argument CPs.We propose both covert movement and focus alternative analysis account for the array of scopal interpretations for in-situ wh-configurations.For future work, we will investigate the issue with other interveners in this language.Additionally, we plan to investigate when interveners result in ungrammaticality in Newari, as opposed to obligatorily embedded scope.Finally, we seek to clarify why verbal-argument CPs are islands for covert movement, and why the patterns observed in Newari are not found in better-studied languages.
A. Extra Data.Examples of wh-in-situ exhibiting no island effects of complex NP construction in non-argument CP clause.In ( 45) and ( 46 Examples of wh-in-situ exhibiting no island effects of 'because'-clause inside of non-argument CP clause, as in (53), the wh-operator 'gana' takes wide scope, while exhibiting island effects in inside of argument CP clause, as it does not take wide scope in (54).Examples of wh-in-situ exhibiting no island effects of 'before/after'-clause inside of non-argument CP clause, as in ( 55), the wh-operator 'chu' takes wide scope, while exhibiting island effects in inside of argument CP clause, as it does not take wide scope in ( 56 pasal-e store-LOC wan-i] go-NONPST dha-u.say-PST 'For which thing x, such that Ram said that before Sayal brings x, Sita will go to the store.' Examples of wh-in-situ exhibiting no island effects of 'wh'-clause inside of non-argument CP clause, as in (57), the wh-operator 'suito' takes wide scope, while exhibiting island effects in inside of argument CP clause, as it does not take wide scope in (58).
place x, such that because I didn't go x, I failed (the test)place x, such that Ram said because I didn't go x, I failed (the test)?' Ram met the teacher y that x hit y?' This offers an account for why neither of the effects are observed in matrix clauses.When FA is blocked by an intervener, the wh-operator can still take the scope via CM, and vice versa.When CM is blocked by an island, FA will become available for the wh-operator to take scope.
).For which person x, such that Qiaofong like the book that Botong wrote to x?' Ram swims more than x runs?' IN VERBAL-ARGUMENT CPS.Generally, argument wh-operators may take sentential scope from a verbal-argument CP: However, unlike matrix clauses or non-argument embedded clauses, we do observe intervention effects in verbal-argument CPs.In (27), we find that the sentential scope interpretation of chu 'what' is blocked.Instead, chu 'what' must be interpreted with embedded scope.Similar findings are demonstrated with (27b) for the adjunct wh-operator chae 'why'.We attribute this obligatory low-scope induced by the addition of the focus operator caka 'only' to an intervention effect.
So far, we have only examined sentences in which the intervener is in the same clause as the wh-operator.Similar patterns obtain when the intervener is in the matrix clause.In (32a), the addition of an intervener results in an ungrammatical sentence.If this results from an intervention effect, then we again predict that moving the clause above the intervener should result in the availability of sentential scope, due to amelioration of the intervention effect.This prediction is confirmed in(33).Moving the most deeply embedded clause (Sitām .chu nau Sita-ERG what eat.PST 'Sita ate what') to the front of the sentence yields sentential scope.
He i said that he j will eat meat.' ), 'chu' and 'guble' take wide scope, and they are grammatical.For which time x, such that Ram know the rumor that Sita ate mango at x time?'Examples of wh-in-situ exhibiting island effects of complex NP construction in argument CP clause.In (47) and (48), 'chu' and 'guble' cannot take wide scope, and they are ungrammatical.For which x, such that Sayal said that Ram know the news that Sita will eat?'When did Akash think Ram know the rumor that Sita ate mango?Examples of wh-in-situ exhibiting no island effects of relative clauses inside of non-argument CP clause, as in (49), the wh-operator 'chu' takes wide scope, while exhibiting island effects in inside of argument CP clause, as it does not take wide scope in (50).For which person x, such that Akash said that Ram met the teacher who x hit?'Examples of wh-in-situ exhibiting no island effects of comparative clause inside of non-argument CP clause, as in (51), the wh-operator 'su-na' takes wide scope, while exhibiting island effects in inside of argument CP clause, as it does not take wide scope in (52).For which person x, such that Akash said Ram swims more than x runs?' ).For which thing x, such that before Sayal brings x, Sita will go to the store.' Sita said Ram is wondering whether Sita hit whom?'