Abstract

This article presents an experiment investigating the relative contribution of two different prosodic properties to the interpretation and scope configuration of wh-indeterminates in Korean. The experiment shows that it is prosodic phrasing after the wh-indeterminate that determines whether it is interpreted as interrogative or indefinite. Prosodic prominence on the wh-indeterminate does not contribute to such a distinction; rather, it increases the possibility of a wide scope reading. The results support the theory that prosodic phrasing is crucial in forming wh-questions, and call for consideration of the influence of prosody on scope-taking properties of wh-indefinites.

1 Introduction

The term wh-indeterminate (see Kuroda 1965) refers to a class of words that can be used to yield interrogative and indefinite readings, as illustrated in the following Korean example:

(1)

  • Ne nwukwu cohaha-ni?1

  • you WH/IND like-Q2

    • a.

      ‘Who do you like?’

    • b.

      ‘Do you like anyone?’

Wh-indeterminates are attested in many languages in the world (Haspelmath 1997), and several crosslinguistic patterns have been observed regarding their properties. One such observation is that the different readings of wh-indeterminates are marked by different prosody: for an interrogative reading, a prosodic domain that spans the wh-indeterminate and the corresponding complementizer is created by removing the prosodic phrase boundaries between them (Richards 2010), while an indefinite reading does not involve such a “dephrasing” effect. Another crosslinguistic observation on wh-indeterminates concerns the relation between their morphology when they are used as indefinites and the possible scope configuration: indefinites that have exactly the same form as interrogatives (henceforth, bare wh-indefinites; e.g., shenme ‘what/something’ in Chinese) are fairly limited in their scope configuration, while indefinites that are marked explicitly by attaching a certain affix to the interrogative form (henceforth, complex wh-indefinites; e.g., nani ‘what’ vs. nani-ka ‘something’ in Japanese) take free scope (Bruening 2007).

Wh-indeterminates in Korean present a challenge to the above two generalizations. First, most of the literature on Korean wh-indeterminates has not identified prosodic phrasing as key to their interpretation; rather, prominence on the wh-word has been noted as the mark of an interrogative reading (e.g., Chang 1973, Choe 1985, Kang 1988, Suh 1989, Kim 2000). Second, there are conflicting judgments on the possible scope configuration of bare wh-indefinites in Korean. While some argue that Korean bare wh-indefinites can only take narrow scope with respect to other scope-taking elements in the sentence (e.g., Ha 2004), my own empirical observations suggest that they can take wide scope, even out of a scope island such as a conditional clause.

In this article, I present an experiment showing that it is prosodic phrasing that distinguishes wh-interrogatives from wh-indefinites in Korean. When not accompanied by dephrasing, prominence on the wh-word does not induce an interrogative reading; rather, it increases the possibility of a wide scope indefinite reading. The experimental results in Korean reinforce the importance of the prosodic domain of wh-questions crosslinguistically and call for consideration of the influence of prosody when studying the semantics of wh-indefinites in other languages.

2 Generalizations regarding Wh-Indeterminates and the Korean Challenge

2.1 Prosodic Disambiguation

Studies on wh-indeterminates have noted that the interrogative and indefinite readings are distinguished by prosody (e.g., Chinese: Hu 2002, Dong 2009; Japanese: Deguchi and Kitagawa 2002, Ishihara 2002). One prosodic property that has been crosslinguistically attested as characterizing wh-interrogatives is prosodic phrasing. Richards (2010:145) argues that “every language tries to create a prosodic structure for wh-questions in which the wh-phrase and the corresponding complementizer are separated by as few prosodic boundaries as possible.” According to him, wh-movement languages achieve this goal by moving the wh-word adjacent to its matching complementizer, whereas wh-in-situ languages do so by deleting prosodic phrase boundaries between the wh-word and the complementizer. He does not develop this argument in the context of disambiguating wh-indeterminates; however, if dephrasing is the characteristic property of interrogatives as he argues, the natural expectation is that phrasing should play a crucial role in disambiguating wh-indeterminates.

(2)

  • Generalization regarding prosody based on Richards 2010 

  • Wh-interrogatives induce prosodic dephrasing up to their corresponding complementizer, while wh-indefinites do not.

Since Korean is a head-final, wh-in-situ language, it is expected to mark its wh-questions by removing prosodic phrase boundaries after the wh-word until the end of the interrogative clause (henceforth, post-wh dephrasing). However, much of the literature has described the prominence (especially high pitch) on wh-indeterminates as the distinctive prosodic property that distinguishes their interrogative use from their indefinite use in Korean (e.g., Chang 1973, Choe 1985, Kang 1988, Suh 1989, Kim 2000). A few scholars have noted that Korean wh-interrogatives also involve a dephrasing effect (Cho 1990, Jun and Oh 1996), but the relative significance of prominence and phrasing is yet to be investigated. Moreover, previous discussions on post-wh dephrasing have been limited to a local effect. Cho (1990:56) argues that wh-interrogatives “form a phonological phrase with the following word,” which does not guarantee a complete prosodic domain between the wh-word and the complementizer. Jun and Oh’s (1996) experimental results indicate that deleting the Accentual Phrase (AP) boundary between a wh-indeterminate and the immediately following word is the most reliably adopted cue to an interrogative reading of the wh-indeterminate in production. However, Jun and Oh’s stimuli were limited to sentences that had only one word after the wh-word, a factor that obscured whether dephrasing includes only the following word or continues to the end of the sentence. This calls for an experimental study to investigate how the presence or absence of global post-wh dephrasing affects the interpretation of wh-indeterminates.

2.2 Scope Configuration

It has long been noticed that indefinites take free scope, even out of syntactic islands such as a conditional clause, as shown in (3) (e.g., May 1985).

(3) John will be happy if someone comes to the party. (if > some, some > if)

However, the scope-taking properties of indefinites with the same form as interrogatives (bare wh-indefinites) are known to be limited (e.g., Chinese: Cheng 1991; Dutch and German: Postma 1994; Russian: Yanovich 2005): only a narrow scope reading (‘if > some’) is available when they appear in conditional clauses as in (3). On the other hand, indefinites that are derived from interrogatives by attaching an affix (complex wh-indefinites) exhibit the same free scope-taking property as regular indefinites. Bruening (2007:159) compares the two types of wh-indefinites in a number of languages and presents the following generalization: “[W]h-indefinites that do not include additional morphology are precluded from taking wide scope (and in fact usually take only narrowest scope), but wh-indefinites that do include additional morphology may take wide scope and may even be interpreted referentially (as specific indefinites).”

(4) Generalization regarding scope in Bruening 2007 

  • a.

    Bare wh-indefinites do not take wide scope.3 (e.g., if > some, *some > if )

  • b.

    Complex wh-indefinites can freely take wide scope. (e.g., if > some, some > if )

Table 1

Judgments on the scope configuration of (5)

Preferred readingsome > ifif > someBothaNeitherbTotal
nwukwu 15 (40.5%) 14 (37.8%) 6 (16.2%) 2 (5.4%) 37 
nwukwu-nka 12 (32.4%) 19 (51.4%) 6 (16.2%) 0 (0.0%) 37 
Preferred readingsome > ifif > someBothaNeitherbTotal
nwukwu 15 (40.5%) 14 (37.8%) 6 (16.2%) 2 (5.4%) 37 
nwukwu-nka 12 (32.4%) 19 (51.4%) 6 (16.2%) 0 (0.0%) 37 

aBoth means that the speaker found the two interpretations equally good.

bNeither indicates that the speaker abstained from making any judgment on (5).

Korean is not discussed in Bruening 2007; however, it provides an interesting test case because it exhibits both types of wh-indefinites (e.g., nwukwu ‘who/someone’, nwukwu-nka ‘someone’). Ha (2004) argues that in Korean, bare wh-indefinites cannot take wide scope whereas complex wh-indefinites can, which coincides with Bruening’s generalization. However, my own judgment and observation suggest that both types of wh-indefinites can take wide scope. For example, I asked 37 linguistically naive Korean speakers to read the sentence in (5) and say whether they interpreted it as ‘Chelswu will be glad if a specific person comes’ or ‘Chelswu will be glad if someone comes (it doesn’t matter who comes)’. The responses summarized in table 1 indicate that a wide scope reading (‘some > if ’) of the bare wh-indefinite was indeed possible and even preferred for many speakers, which is an unexpected result according to Ha (2004) and Bruening (2007). This calls for a controlled experiment to investigate the possible scope configurations of bare wh-indefinites.

(5)

  • Nwukwu(-nka)-ka o-myen Chelswu-ka cohaha-lke-ta.

  • WH/IND(-IND)-NOM come-if Chelswu-NOM glad-will-DCL

  • ‘Chelswu will be glad if someone comes.’

  • (cf. Ha 2004:92)

3 Experiment

The questions from section 2 can be summarized as follows: (a) Which prosodic factor is crucial in deciding the meaning of the wh-indeterminates: prominence or dephrasing? (b) Can bare wh-indefinites take wide scope? Why are there different judgments on the scope configuration of wh-indefinites? In the rest of the article, I will show that the answers to these questions are interrelated. Let us first consider regular indefinite expressions such as some in English. It has long been reported that stressed and unstressed some have different semantic properties that can affect their scopal behavior (e.g., Milsark 1974, Lohndal 2010), supporting the analysis that stress correlates with a wide scope reading of regular indefinites. Extending this line of analysis from regular indefinites to wh-indefinites, I propose the following hypotheses:

(6) Hypotheses

  • a.

    Prosodic phrasing determines the meaning of wh-indeterminates.

  • b.

    Prosodic prominence affects the scope configuration of wh-indeterminates.

Table 2

Factors crossed in design of stimuli

Wh-pitch raising
No raising (r)Raising (R)
Post-wh dephrasing No dephrasing (d) rd Rd 
Dephrasing (D) rD RD 
Wh-pitch raising
No raising (r)Raising (R)
Post-wh dephrasing No dephrasing (d) rd Rd 
Dephrasing (D) rD RD 

Then, conflicting judgments on the scope configuration of wh-indefinites can be attributed to the lack of consideration of prosody (see Fodor 2002). The rest of this section introduces an experiment to test the hypotheses.

3.1 Overview and Predictions

According to the literature on the prosody of Korean wh-indeterminates cited in section 2.1, wh-interrogative prosody involves both pitch raising on a wh-indeterminate and dephrasing after it, whereas neither is found in wh-indefinite prosody. Since these two factors (wh-pitch raising and post-wh dephrasing) pattern together in natural speech, I detached them and created synthesized stimuli to assess their relative contribution. The prosodic factors in the stimuli were manipulated to create a 2 × 2 design as shown in table 2, where an uppercase acronym indicates the presence of a property and a lowercase acronym indicates its absence. The [rd] contour indicates canonical indefinite prosody, while [RD] indicates canonical interrogative prosody. The main purpose of the experiment was to see how listeners interpret the noncanonical cases [rD] and [Rd]. Hypothesis (6a) predicts that [rD] must be interpreted as a wh-question more often than [Rd]. If wh-pitch raising is the more important factor, on the other hand, [Rd] must be interpreted as a wh-question more often than [rD]. In addition, the experiment investigated whether scope configuration can be affected by prosody. Hypothesis (6b) predicts that the stimuli with pitch raising on the wh-word (coded as “R”) will be perceived as having a wide scope reading more often than the stimuli without such pitch raising (coded as “r”). The hypothesis should be rejected if there is no difference in perceiving scope configuration whether there is wh-pitch raising or not.

3.2 Method

3.2.1 Stimuli

An example stimulus is shown in (7). A neutral sentence ending was used to render the sentence ambiguous between an assertion and a question, and a wh-indeterminate phrase was placed in a conditional clause. Thus, three different readings of the sentence were available: a declarative sentence with a narrow scope indefinite or a wide scope indefinite, or a wh-question.4 In theory, a yes/no question reading was also available; however, that possibility was ruled out in the experiment because all the stimuli were constructed to have falling intonation at the end of the sentence, while sharp rising sentence-final intonation (H%) is a distinctive property of yes/no questions that contain wh-indefinites (Lee 1997).

Figure 1

Pitch tracks of two different readings of the sentence (7).

The annotation follows the K-ToBI convention (Jun 2000), but for simplicity only surface tones are marked here. For the notation of tones, let T indicate L (Low) or H (High) or their combination. Then T: AP-initial tone; +T, T+: AP-medial tone; Ta: AP-final tone; T%: IP boundary tone. The AP-final tone in the sentence-final AP (Aspectual Phrase) is overridden by the IP (Intonation Phrase) boundary tone.

Figure 1

Pitch tracks of two different readings of the sentence (7).

The annotation follows the K-ToBI convention (Jun 2000), but for simplicity only surface tones are marked here. For the notation of tones, let T indicate L (Low) or H (High) or their combination. Then T: AP-initial tone; +T, T+: AP-medial tone; Ta: AP-final tone; T%: IP boundary tone. The AP-final tone in the sentence-final AP (Aspectual Phrase) is overridden by the IP (Intonation Phrase) boundary tone.

(7)

  • I yak-ey mwe-ka tuleka-myen wihemhay

  • this reagent-LOCWH/IND-NOM get.into-if dangerous.DCL/Q

  • ‘It is dangerous if something gets into this reagent.’ (‘if > some’, ‘some > if ’)

  • or ‘What is the thing such that it is dangerous if it gets into this reagent?’

Stimuli were 12 sentences similar in structure to (7). (For the complete list, see the online appendix at https://www.mitpressjournals.org/doi/suppl/10.1162/ling_a_s00318.) They were recorded by a female native speaker of Seoul Korean in her twenties, who had training in linguistics in college. The speaker read written sentences in two different settings. In the first setting, the sentences were presented with a period at the end to facilitate a declarative sentence interpretation. No further context was provided in this setting, in order to induce an utterance that bore no focus on any particular item. In the second setting, the sentences were presented with a question mark at the end and followed by an answer that facilitated a wh-question interpretation. The recording was made in the sound-attenuated booth of the Phonetics Lab at Seoul National University.

Figure 1 presents the pitch tracks of the sentence (7). The two readings exhibited prosodic differences in terms of both wh-pitch raising and post-wh dephrasing across all recorded pairs of sentences.5 Wh-pitch raising: For the wh-interrogative reading, the highest pitch point in the sentence was observed within the wh-phrase; for the declarative reading, it was observed outside the wh-phrase (specifically, on the morpheme -myen ‘if ’). The highest pitch value in the wh-region was also different for the two readings (declarative: 250.3 Hz; wh-question: 278.9 Hz on average; paired t-test: p < .001).6Post-wh dephrasing: In the declarative reading, each word consisted of one AP (see figure 1a); in the wh-question reading, the wh-word and all following words were in the same, large AP, showing fewer AP tones and smoother pitch contour after the wh-word (see figure 1b).7

To create the actual stimuli used in the experiment, the recording of declarative sentences was chosen as the base and manipulated with Praat (Boersma 2001), as illustrated in figure 2.8 The first type of stimulus (figure 2a) was created by stylizing the pitch contours to points that represented AP tonal targets (Jun 2005). It was supposed to maintain the overall shape of the base contour. The second type of stimulus (figure 2b) was obtained by raising the highest pitch of the wh-indeterminate phrase to the same point as the highest pitch value in the corresponding wh-question recorded by the speaker. As a result, the highest pitch point of the entire sentence fell on the wh-indeterminate phrase. While this stimulus type was created to replicate the effect of pitch raising on the wh-phrase, a slight amount of manipulation in the post-wh region was added: the immediate post-wh L tone was moved to a delayed position (i.e., the penultimate syllable of the post-wh word) to avoid sharp falling after the pitch peak in the wh-indeterminate.9 The third type of stimulus (figure 2c) was obtained by erasing pitch points after the wh-phrase up to the penultimate syllable of the sentence. It was created to replicate the effect of global post-wh dephrasing that deletes post-wh AP tones (see Jun 1993 for dephrasing in terms of deleting AP tones). The fourth type of stimulus (figure 2d) was obtained by applying both wh-pitch raising and post-wh dephrasing as described above. It was created to replicate the canonical intonation pattern of wh-questions. The pitch points associated with the last syllable of the sentence remained unchanged across all stimuli; thus, the sentence-final intonation was always kept the same.

Figure 2

Pitch tracks of auditory stimuli based on the sentence in (7).

The intersection point of the two guidelines indicates the point of highest pitch in the wh-indeterminate phrase.

Figure 2

Pitch tracks of auditory stimuli based on the sentence in (7).

The intersection point of the two guidelines indicates the point of highest pitch in the wh-indeterminate phrase.

3.2.2 Participants

Participants in the experiment were adult native speakers of Korean (N = 57, age > 18) who had lived more than 10 years in Seoul or the vicinity where Seoul Korean is spoken. They were recruited online through various social networking services and volunteered their time (15 minutes on average) without payment. Participation in the experiment was anonymous, but the source of recruitment suggests that participants were mostly college students in Korea who were different from participants in the informal survey described in section 2.2.

3.2.3 Procedure

A total of 48 stimuli (12 sentences × 4 intonation types) were created through manipulation, and they were divided into four sets so that each set included only one version of each of the 12 sentences to avoid presenting the same strings repeatedly.10 For each set, the 12 target stimuli were arranged in a pseudorandom order and intermingled with 23 filler stimuli. The fillers were sentences containing wh-indeterminates in various constructions other than conditional clauses. The filler materials were recorded by the same speaker who produced the base of the target materials. Five filler sentences were presented at the beginning of the experiment as a training session.

The experiment was delivered through Qualtrics (online survey software). All materials were presented in Korean. Using the randomizer of Qualtrics with the Evenly Present Elements option, each participant was randomly but evenly assigned to one of the four sets of stimuli. The stimuli were presented in a self-paced forced-choice task in the following way. For each stimulus, a screen displayed four interactive elements: a button to play the sound of the stimulus, two choice forms to elicit responses, and a button to move to the next stimulus. The participants first pressed the Play button to listen to the stimulus, delivered through headphones. The stimulus was provided only aurally; no text was given. Then the participants pressed an option button in the first choice form that was associated with either Question or Statement to indicate their interpretation of the stimulus. The second choice form that the participants saw depended on their response to the first choice form. If they chose Question in the first form, in the second form they were asked to choose whether it was a yes/no question or a wh-question. If they chose Statement in the first form, in the second form they were asked to choose whether it was about a specific entity (i.e., a wide scope indefinite) or an arbitrary entity (i.e., a narrow scope indefinite). The participants were allowed to listen to the stimulus repeatedly and change their responses freely at any time until they moved to the next stimulus. Figure 3 illustrates the flow of the task, taking the sentence in (7) as an example.

As shown in figure 3, the task was in fact choosing among four readings (i.e., yes/no question, wh-question, statement with a wide scope indefinite, and statement with a narrow scope indefinite). Instead of presenting the four choices at once, however, I divided them into two groups according to their illocutionary force and added an intermediate step of choosing between the two groups so that the participants always made a binary choice. I did this to reduce the cognitive load during the experiment and prevent dropout due to the complicated nature of the task.

Since (as mentioned earlier) the target stimuli were not expected to receive a yes/no question response because of their final falling intonation, the filler stimuli included nine yes/no questions to balance the response matrix. The yes/no question fillers also served to qualify participants and detect outliers whose responses were suspected to be unreliable (see Cowart 1997): all participants chose the intended reading of yes/no question stimuli quite consistently (90% of the time on average), except for four participants who recognized yes/no questions less than 50% of the time. The responses from those four participants were excluded from the analysis.11

3.3 Results and Analysis

Table 3 presents the number of responses for each intonation type.12 As expected, yes/no question responses were very infrequent (3 out of 636). Null responses were also rare (4 out of 636) and are excluded from the analyses below.

Figure 3

Flow of the forced-choice task.

(YNQ = yes/no question; WHQ = wh-question; WS = wide scope; NS = narrow scope)

Figure 3

Flow of the forced-choice task.

(YNQ = yes/no question; WHQ = wh-question; WS = wide scope; NS = narrow scope)

Table 3

Number of responses for each intonation type

StimuliResponses
Declarative: Narrow scope indefiniteDeclarative: Wide scope indefiniteWh-questionYes/no questionNo responseTotal
[rd] 114 39 157 
Base contour       
[Rd] 74 70 16 162 
Wh-pitch raising       
[rD] 29 25 105 160 
Post-wh dephrasing       
[RD] 23 34 96 157 
Wh-pitch raising + post-wh dephrasing       
Total 240 168 221 636 
StimuliResponses
Declarative: Narrow scope indefiniteDeclarative: Wide scope indefiniteWh-questionYes/no questionNo responseTotal
[rd] 114 39 157 
Base contour       
[Rd] 74 70 16 162 
Wh-pitch raising       
[rD] 29 25 105 160 
Post-wh dephrasing       
[RD] 23 34 96 157 
Wh-pitch raising + post-wh dephrasing       
Total 240 168 221 636 

3.3.1 Meaning of Wh-Indeterminates (Interrogative vs. Indefinite)

Since all responses except for a wh-question response indicate that the wh-indeterminates in the stimuli were perceived as indefinite, the proportion of wh-question responses was chosen to be presented in figure 4 to indicate the relation between the intonation type and the meaning of wh-indeterminates. The base [rd] contour, which had almost the same contour as a declarative sentence, was interpreted as a wh-question only 3% of the time (4 out of 157). On the other hand, the [RD] contour, which was obtained by manipulating the base contour to bear two canonical properties of wh-questions (i.e., wh-pitch raising and post-wh dephrasing), was interpreted as a wh-question 62% of the time (96 out of 156). When it comes to the stimuli to which only one property of wh-questions was added, the [Rd] contour, which had wh-pitch raising but lacked post-wh dephrasing, was interpreted as a wh-question only 10% of the time (16 out of 160). The [rD] contour, which had post-wh dephrasing but lacked wh-pitch raising, was interpreted as a wh-question 66% of the time (105 out of 159), which is compatible with the results of the [RD] contour. These results are consistent with the hypothesis that post-wh dephrasing is the crucial factor in deciding the meaning of wh-indeterminates, while wh-pitch raising is not.

To assess the statistical strength of the prosodic effects, a logistic mixed-effects model was employed in R (R Core Team 2015), using the glmer function from the lme4 package (Bates et al. 2014). The model predicted wh-question response (WHQ: 1; others: 0) with ( pitch) raising, dephrasing, and their interaction as fixed effects (R: 0.5; r: −0.5; D: 0.5; d: −0.5). The model also included random intercepts for subject and item as well as random slopes for raising by subject, which was the maximal random-effects structure justified by the data (Baayen, Davidson, and Bates 2008). Table 4 presents the result of the logistic regression. The model confirms that dephrasing predicted wh-question response ( p < .001), while there was no main effect of raising ( p = .142). Rather, the interaction between raising and dephrasing reached statistical significance ( p < .022), which indicates that raising lowered the probability of wh-question interpretations when it combined with dephrasing.13

Figure 4

Proportion of wh-question responses for each intonation type

Figure 4

Proportion of wh-question responses for each intonation type

Table 4

Logistic mixed-effects model on wh-question responses

PredictorβSE(β)zp > |z|
(Intercept−1.57 0.40 −3.91  
Raising 0.53 0.36 1.47 .142 
Dephrasing 4.77 0.43 11.01 <.001 
Raising: dephrasing −1.77 0.77 −2.30 .022 
PredictorβSE(β)zp > |z|
(Intercept−1.57 0.40 −3.91  
Raising 0.53 0.36 1.47 .142 
Dephrasing 4.77 0.43 11.01 <.001 
Raising: dephrasing −1.77 0.77 −2.30 .022 

3.3.2 Scope of Wh-Indefinites (Wide vs. Narrow)

To assess the effect of intonation on the scope configuration of wh-indefinites, the declarative responses were separated out; their scope readings are presented in table 5. The base [rd] contour received a wide scope indefinite response 25% of the time (39 out of 153), whereas the wide scope ratio of the wh-prominent [Rd] contour was considerably higher, 49% (70 out of 144). A similar positive correlation between wh-pitch raising and wide scope interpretation was observed for the stimuli involving dephrasing: the wide scope ratio was 46% (25 out of 54) for the [rD] contour but 60% (34 out of 57) for the [RD] contour. Figure 5 presents the proportion of wide scope indefinite responses out of the declarative responses.

Table 6 presents the estimation of a logistic mixed-effects model, which predicted wide scope response (WS: 1; NS: 0) with the same fixed and random effects as the previous model.14 The model confirms that there was a significant and positive main effect of raising ( p < .006). The main effect of dephrasing was also significant and positive ( p < .001), while the interaction of raising and dephrasing was not statistically significant ( p =.082).15

Table 5

Number of declarative responses for each intonation type

ResponseStimuli
[rd][Rd][rD][RD]Total
Narrow scope 114 74 29 23 240 
Wide scope 39 70 25 34 168 
Total 153 144 54 57 408 
ResponseStimuli
[rd][Rd][rD][RD]Total
Narrow scope 114 74 29 23 240 
Wide scope 39 70 25 34 168 
Total 153 144 54 57 408 
Figure 5

Proportion of wide scope responses out of declarative responses for each intonation type

Figure 5

Proportion of wide scope responses out of declarative responses for each intonation type

Table 6

Logistic mixed-effects model on wide scope indefinite responses

PredictorβSE(β)zp > |z|
(Intercept−0.12 0.45 −0.26  
Raising 1.01 0.37 2.76 .006 
Dephrasing 1.32 0.35 3.80 <.001 
Raising:dephrasing −1.10 0.63 −1.74 .082 
PredictorβSE(β)zp > |z|
(Intercept−0.12 0.45 −0.26  
Raising 1.01 0.37 2.76 .006 
Dephrasing 1.32 0.35 3.80 <.001 
Raising:dephrasing −1.10 0.63 −1.74 .082 

4 Discussion

4.1 Post-Wh Dephrasing

While a typical wh-question contour bears both wh-pitch raising and post-wh dephrasing, the experimental results indicate that it is only dephrasing that contributes to the perception of a wh-question. The stimuli in which the wh-indeterminate was boosted but not followed by dephrasing until the end of the sentence were interpreted as wh-questions only 10% of the time. On the other hand, the stimuli involving post-wh dephrasing received wh-question interpretations more than 50% of the time, whether they involved wh-prominence or not.16

The experimental results also suggest that post-wh dephrasing should be global. Recall that the [Rd] contour stimuli involved not only an expanded pitch range on the wh-word but also a slight deviation from the base in the post-wh region (i.e., delayed appearance of the post-wh L tone). This additional manipulation amounts to a local dephrasing effect because the post-wh L tone could have marked the beginning of a separate AP if it had not been delayed. However, the stimuli manipulated in this way were still interpreted as indefinites 90% of the time. In other words, the results suggest that post-wh dephrasing should continue up to the complementizer.

If prominence does not contribute to identifying wh-interrogatives in perception, why does typical wh-question prosody involve pitch raising of the wh-word in production? If we accept the argument that a wh-indeterminate word receives the focus of the sentence when it is used as an interrogative (Deguchi and Kitagawa 2002), wh-interrogatives are likely to share a prosodic characteristic of focus. Considering the crosslinguistic observation that expanded pitch range is a common property of focus prosody (Flemming 2008), we can think of the boosted pitch on wh-interrogatives as an indicator of their focus feature rather than their interrogative feature. Interestingly, it has been observed that focus also induces a dephrasing effect, to a certain degree (Jun and Lee 1998). However, the status of dephrasing in focus prosody seems auxiliary rather than crucial, unlike in wh-interrogative prosody, since post-wh dephrasing is consistently observed (Jun and Oh 1996) but postfocus dephrasing is optional (Jun and Lee 1998) in Korean.

The apparent similarities of focus prosody and wh-prosody have also been observed in Japanese (e.g., Deguchi and Kitagawa 2002, Ishihara 2002), where both involve boosting the pitch on the target word and removing the lexical accents on the following words. However, Sugahara (2005) shows that prosodic phrase boundaries are still observed after a focused item, which suggests that dephrasing is not a necessary property in focus prosody, but rather a side effect of postfocus pitch range compression. In sum, it seems that focus prosody involves a gradual increase in pitch range on the focused item, whereas wh-prosody involves the categorical change of deleting post-wh AP tones (in Korean) or post-wh lexical accents (in Japanese).17 Further investigation of the interaction between focus prosody and wh-prosody in production remains a task for future research.

4.2 Prominence of Wh-Indefinites

The experimental results show that prosody can affect the scope configuration of wh-indefinites in Korean. For the base stimuli that were obtained by recording declarative sentences without particular prominence on any word, 75% of the time, participants who made a declarative response then indicated a narrow scope indefinite reading (114 out of 153). This suggests that there is a baseline preference for a narrow scope reading for bare wh-indefinites in Korean when they are not prosodically prominent. This may explain Ha’s (2004) impressionistic observation that Korean bare wh-indefinites do not take wide scope, since a default prosodic contour tends to influence the syntactic judgment (Fodor 2002). Yet the results also indicate that 25% of the time, participants preferred a wide scope indefinite reading even without prominence on the wh-word. Furthermore, when the wh-word was boosted, the wide scope response rate became significantly higher (49%; 70 out of 144).18 Thus, we can conclude that a wide scope reading is available for Korean bare wh-indefinites, and that prominence on the wh-word increases the possibility of a wide scope reading.

What is the crosslinguistic implication of the above conclusion? Recall that many languages have been argued to restrict their bare wh-indefinites to a narrow scope interpretation (e.g., Bruening 2007). The experimental results suggest that prosodic prominence can have an effect similar to that of a morphological affix in deciding scope configuration in that they both enable wh-indefinites to take scope freely. If there are languages that never allow a wide scope reading of bare wh-indefinites, it might be the case that such languages have a prosodic constraint that bare wh-indefinites are never realized as prominent. Indeed, bare wh-indefinites in many languages have been reported to be prosodically unmarked or reduced (e.g., Classical Greek: Haspelmath 1997; Mandarin Chinese: Hu 2002, Dong 2009). More evidence for the correlation between prosodic prominence and the scope configuration of wh-indefinites is likely to be found in other languages.

5 Conclusion

This article has provided empirical and experimental observations in support of the following two arguments. First, in deciding the meaning of a wh-indeterminate in Korean, post-wh dephrasing is a crucial factor while pitch raising on the wh-word is not. The experimental results in this article coincide with Cho’s (1990) and Jun and Oh’s (1996) argument that wh-questions crucially involve dephrasing after the wh-word in Korean, and further suggest that the dephrasing effect should be a global one that continues up to the corresponding question complementizer. The results also support the theory that creating a prosodic domain between a wh-word and the corresponding complementizer is crosslinguistically crucial in forming wh-questions (Richards 2010).

Second, Korean bare wh-indefinites can take wide scope even out of a scope island, and the island-escaping property is further strengthened when the wh-indefinite receives prosodic prominence. The experimental results urge reconsideration of the crosslinguistic generalization that bare wh-indefinites cannot take wide scope (Bruening 2007) and call for experimental studies in other languages to investigate whether their bare wh-indefinites can receive prosodic prominence, and if so, whether prosodic prominence enhances a wide scope reading.

Notes

1 A modified Yale system (cf. Martin 1992) with a strict one-to-one mapping between Korean and Roman scripts is adopted in this article for transliterating Korean data, to clearly indicate morphosyntactic units. In the annotation of sound files, phonetic transcriptions based on the IPA are used.

2 The following abbreviations are used in this article: ACC = accusative case, DCL = declarative sentence ending, DCL/Q = neutral sentence ending, LOC = locative case, IND = indefinite marker, NOM = nominative case, TOP = topic, Q = question ending, WH/IND = wh-indeterminate.

3 The term wide scope in Bruening 2007 should be interpreted as ‘widest scope’ when there are more than two scope-bearing elements in the sentence because an intermediate scope reading is possible for bare wh-indefinites (Lin 2004).

4 A wh-question reading was available because a conditional clause is not a wh-island in Korean.

5 Two additional differences are observed in figure 1: The pre-wh words belong to separate APs in (a) but the same AP in (b), and the sentence-final tone is L% in (a) but LH% in (b). These differences were ignored in designing the perception experiment because the difference in the pre-wh region did not consistently appear in the other stimulus sentences, and because the choice of sentence-final intonation is known to be a tendency and not a decisive factor for identifying the type of sentence, except for yes/no questions (Lee 1997). Furthermore, an experimental study reported in Yun 2015 suggests that the influence of sentence-final intonation is not as strong as that of prosodic phrasing in perceiving wh-questions.

6 There was no statistically significant difference between the two readings in the wh-region in terms of duration or amplitude.

7 An AP is manifested by a sequence of tones, THLH (T for L or H) (Jun 1998). All four tones can be fully realized if the AP consists of four or more syllables; otherwise, intermediate tones tend not to be realized on the surface.

8Wh-question recordings were used only as a standard for manipulation and not as the actual base of manipulation because creating tonal targets to replicate the declarative prosody is more difficult and has a more unnatural-sounding result than removing tonal targets to replicate the wh-question prosody.

9 I had conducted a pilot study in which this additional manipulation was not employed, and some participants reported that the sharp falling after wh-pitch raising made the sentence sound unnatural. Apparently, this is because when pitch raising occurs in natural speech due to focus, it is usually followed by a certain degree of pitch smoothing (Jun and Lee 1998). The additional manipulation produces an effect similar to local post-wh dephrasing (i.e., a dephrasing effect in the immediate post-wh word only); the implication of this subtle prosodic requirement will be discussed in section 4.1.

10 The contours for the stimuli in each set were counterbalanced using a modified Latin square to make all four contours appear in each set relatively evenly but not exactly the same number of times.

graphic

11 The overall response patterns of those four participants were indeed quite different from the general pattern. Two participants never chose questions in their responses, and a third never chose statements. The response pattern of the fourth was rather arbitrary.

12 The total number of responses was similar across each intonation type but not exactly the same because the number of participants was not the same for each set (Set 1: 14; Set 2: 15; Set 3: 13; Set 4: 11).

13 This interaction reflects the rather unexpected result that the WHQ-response rate for the [RD] contour was lower than that for the [rD] contour. Another perception study involving synthesized speech in Korean (Yun and Lee to appear) reports a similar case where wh-indeterminates were interpreted as interrogatives more often when the overall pitch contour of the sentence was closer to a straight line. Since a straight contour implies no phrasing boundaries and no prominence in terms of pitch, this result provides another piece of evidence for the argument that what is crucial for a wh-question reading is dephrasing, not pitch raising.

14 The same random effects were proved to be the maximal random effects structure by a series of likelihood tests.

15 The effect of dephrasing on scope configuration was not part of the hypothesis in (6). A possible explanation (Michael Wagner, pers. comm.) is that post-wh dephrasing could enhance the prominence of the wh-indeterminate because if all the post-wh words lost their AP tones due to dephrasing, the wh-word would become perceptually more prominent even if it did not receive high pitch (see figure 2c). If this is the case, the higher wide scope response rate for the stimuli with dephrasing would still be attributed to the prominence of the wh-indeterminate.

16 The percentage of wh-question responses to all dephrasing stimuli was 64% (201 out of 315), which is higher than 50% yet far lower than 100%. Possible reasons why dephrasing did not induce more wh-question responses include the following: First, manipulating only the pitch contour might have left other phrasing cues to indefinites in the base recording (such as voicing and segment quality; see Jun 1993). Second, the L% sentence-final intonation was retained for all stimuli to simplify the experimental design, but the most frequent sentence-final tone for wh-questions is LH% (Jun and Oh 1996). Although L% is possible for wh-questions (Jun and Oh 1996, Lee 1997), the use of this noncanonical tone could have lowered the wh-question response rate.

17 A dialectal variation also supports the argument that focus prosody and wh-interrogative prosody are not the same. Hwang (2011) notes that in Kyungsang Korean, postfocus prosody is realized as pitch range reduction while post-wh prosody is realized as high plateau.

18 The narrow scope response ratio for the wh-raised stimuli was still high (51%); however, this does not mean that the participants who made those responses considered that a wide scope indefinite reading was impossible, since it was a forced-choice task asking participants to indicate a preferred reading.

Acknowledgments

I would like to thank Mark Aronoff, Michael Becker, Jason Bishop, Laura Casasanto, Masayuki Gibson, Jonathan Howell, Sun-Ah Jun, Seongyeon Ko, Nayoung Kwon, Richard Larson, Lori Repetti, Norvin Richards, Mats Rooth, Michael Wagner, and John Whitman for helpful discussions and valuable comments on this work at different stages. I also thank Hyunah Baek, Semoon Hoe, Yoon-Shin Kim, Heejeong Ko, Saera Kwak, Hoyoung Lee, So Young Lee, and Yoon-Kyung Song for helping with recruitment for the experiment, and I thank all the participants in this study for sharing their valuable time. I appreciate the Qualtrics support team for their timely technical help. Earlier versions of this work were presented at WCCFL 29, Harvard-ISOKL 14, Stony Brook University, Seoul National University, Montclair State University, Konkuk University, and CUNY Graduate Center, and I am grateful to the audiences for their helpful feedback. Finally, I thank the anonymous referees for their constructive criticism, which has improved this article. All remaining errors are mine.

References

References
Baayen,
R. Harald
,
Douglas J.
Davidson
, and
Douglas M.
Bates
.
2008
.
Mixed-effects modeling with crossed random effects for subjects and items
.
Journal of Memory and Language
59
:
390
412
.
Bates,
Douglas
,
Martin
Maechler
,
Ben
Bolker
, and
Steven
Walker
.
2014
.
lme4: Linear mixed-effects models using Eigen and S4. R package version 1
.
1
7
. http://CRAN.R-project.org/package=lme4.
Boersma,
Paul
.
2001
.
Praat, a system for doing phonetics by computer
.
Glot International
5
:
341
345
.
Bruening,
Benjamin
.
2007
.
Wh-in-situ does not correlate with wh-indefinites or question particles
.
Linguistic Inquiry
38
:
139
166
.
Chang,
Suk-Jin
.
1973
. A generative study of discourse: Pragmatic aspects of Korean with reference to English.
Ehak yenkwu [Language Research] 9.2 (supplement)
.
Seoul
:
Seoul National University, Language Research Institute
.
Cheng,
Lisa Lai-Shen
.
1991
.
On the typology of wh-questions
.
Doctoral dissertation, MIT, Cambridge, MA
.
Cho,
Young-mee Yu
.
1990
. Syntax and phrasing in Korean. In
The phonology-syntax connection
, ed. by
Sharon
Inkelas
and
Draga
Zec
,
47
62
.
Chicago
:
University of Chicago Press
.
Choe,
Jae-Woong
.
1985
.
Pitch-accent and q/wh words in Korean
.
Harvard Studies in Korean Linguistics
1
:
113
123
.
Cowart,
Wayne
.
1997
.
Experimental syntax: Applying objective methods to sentence judgments
.
Thousand Oaks, CA
:
Sage Publications
.
Deguchi,
Masanori
, and
Yoshihisa
Kitagawa
.
2002
.
Prosody and wh-questions
. In
NELS 32
, ed. by
Mako
Hirotani
,
73
92
.
Amherst
:
University of Massachusetts, Graduate Linguistic Student Association
.
Dong,
Hongyuan
.
2009
.
Issues in the semantics of Mandarin questions
.
Doctoral dissertation, Cornell University, Ithaca, NY
.
Flemming,
Edward
.
2008
.
The role of pitch range in focus marking
.
Paper presented at
the Workshop on Information Structure and Prosody, Studiecentrum Soeterbeeck
,
Ravenstein, Netherlands
,
21–22 January
.
Fodor,
Janet Dean
.
2002
. Prosodic disambiguation in silent reading. In
NELS 32
, ed. by
Mako
Hirotani
,
113
132
.
Amherst
:
University of Massachusetts, Graduate Linguistic Student Association
.
Ha,
Seungwan
.
2004
. The existential reading of wh-words and their scope relations. In
Proceedings from the Annual Meeting of the Chicago Linguistic Society 40
, ed. by
Nikki
Adams
,
Adam
Cooper
,
Fey
Parrill
, and
Thomas
Weir
,
83
95
.
Chicago
:
University of Chicago, Chicago Linguistic Society
.
Haspelmath,
Martin
.
1997
.
Indefinite pronouns
.
Oxford
:
Clarendon Press
.
Hu,
Fang
.
2002
.
A prosodic analysis of wh-words in Standard Chinese
.
Paper presented at
Speech Prosody 2002
,
Aix-en-Provence, France
,
11–13 April
.
Hwang,
Hyun Kyung
.
2011
.
Scope, prosody, and pitch accent: The prosodic marking of wh-scope in two varieties of Japanese and South Kyeongsang Korean
.
Doctoral dissertation, Cornell University, Ithaca, NY
.
Ishihara,
Shinichiro
.
2002
. Invisible but audible wh-scope marking: Wh-constructions and deaccenting in Japanese. In
WCCFL 21
, ed. by
Line
Mikkelsen
and
Christopher
Potts
,
180
193
.
Somerville, MA
:
Cascadilla Press
.
Jun,
Sun-Ah
.
1993
.
The phonetics and phonology of Korean prosody
.
Doctoral dissertation, Ohio State University, Columbus
.
Jun,
Sun-Ah
.
1998
.
The Accentual Phrase in the Korean prosodic hierarchy
.
Phonology
15
:
189
226
.
Jun,
Sun-Ah
.
2000
.
K-ToBI (Korean ToBI) labelling conventions: Version 3
.
Speech Sciences
7
:
143
169
.
Jun,
Sun-Ah
.
2005
.
Korean intonational phonology and prosodic transcription
. In
Prosodic typology: The phonology of intonation and phrasing
, ed. by
Sun-Ah
Jun
,
201
229
.
Oxford
:
Oxford University Press
.
Jun,
Sun-Ah
, and
Hyuck-Joon
Lee
.
1998
. Phonetic and phonological markers of Contrastive Focus in Korean. In
International Conference on Spoken Language Processing (ICSLP) 5
, ed. by
Robert H.
Mannell
and
Jordi
Robert-Ribes
,
1295
1298
.
Canberra City
:
Australian Speech Science and Technology Association
.
Jun,
Sun-Ah
, and
Mira
Oh,
.
1996
.
A prosodic analysis of three types of wh-phrases in Korean
.
Language and Speech
39
:
37
61
.
Kang,
Myung-Yoon
.
1988
.
Topics in Korean syntax: Phrase structure, variable binding and movement
.
Doctoral dissertation, MIT, Cambridge, MA
.
Kim,
Ae-Ryung
.
2000
.
A derivational quantification of “WH-phrase.”
Doctoral dissertation, Indiana University, Bloomington
.
Kuroda,
S.-Y
.
1965
.
Generative grammatical studies in the Japanese language
.
Doctoral dissertation, MIT, Cambridge, MA
.
Lee,
Ho-Young
.
1997
.
Kwukewunyullon [Korean prosody]
.
Seoul
:
Hankwukyenkwuwen [Korean Study Institute]
.
Lin,
Jo-Wang
.
2004
.
Choice functions and scope of existential polarity wh-phrases in Mandarin Chinese
.
Linguistics and Philosophy
27
:
451
491
.
Lohndal,
Terje
.
2010
.
More on scope illusions
.
Journal of Semantics
27
:
399
407
.
Martin,
Samuel E
.
1992
.
A reference grammar of Korean
.
Rutland, VT
:
Tuttle
.
May,
Robert
.
1985
.
Logical Form: Its structure and derivation
.
Cambridge, MA
:
MIT Press
.
Milsark,
Gary
.
1974
.
Existential sentences in English
.
Doctoral dissertation, MIT, Cambridge, MA
.
Postma,
Gertjan
.
1994
. The indefinite reading of WH. In
Linguistics in the Netherlands 1994
, ed. by
Reineke
Bok-Bennema
and
Crit
Cremers
,
187
198
.
Amsterdam
:
John Benjamins
.
R Core Team
.
2015
.
R: A language and environment for statistical computing
.
Vienna
:
R Foundation for Statistical Computing
. http://www.R-project.org/.
Richards,
Norvin
.
2010
.
Uttering trees
.
Cambridge, MA
:
MIT Press
.
Sugahara,
Mariko
.
2005
.
Post-focus prosodic phrase boundaries in Tokyo Japanese: Asymmetric behavior of an F0 cue and domain-final lengthening
.
Studia Linguistica
59
:
144
173
.
Suh,
Cheong-Soo
.
1989
.
Interrogatives and indefinite words in Korean: With reference to Japanese
.
Harvard Studies in Korean Linguistics
3
:
329
340
.
Yanovich,
Igor
.
2005
. Choice-functional series of indefinites and Hamblin semantics. In
SALT 15
, ed. by
Efthymia
Georgala
and
Jonathan
Howell
,
309
326
. https://journals.linguisticsociety.org/procee dings/index.php/SALT/issue/view/94.
Yun,
Jiwon
.
2015
. The influence of sentence-final intonation and phonological phrasing on the interpretation of wh-indeterminates. In
Proceedings of the 9th Workshop on Altaic Formal Linguistics
, ed. by
Andrew
Joseph
and
Esra
Predolac
,
25
34
.
MIT Working Papers in Linguistics 76
.
Cambridge, MA
:
MIT, MIT Working Papers in Linguistics
.
Yun,
Jiwon
, and
Hye-Sook
Lee,
. To appear. Prosodic disambiguation of questions in Korean: Theory and processing.
Korean Linguistics
.