Production and Interpretation of Answers to Negative Polar
Questions in Dutch
Ariska Iris Bonnema
MA thesis
Department of Applied Linguistics Faculty of Arts, University of Groningen
Student number: s2366150
Supervisor: Prof. Dr. Petra Hendriks Second reader: Prof. Dr. Wa nder Lowie Completion date: June 19, 2020
Abstract
Responses to negative polar questions are ambiguous. When answering a negative polar question in English (e.g., Don’t you want coffee?) both yes and no can be used to respond affirmatively and negatively to the question. Dutch, compared to English, has an extra polarity particle jawel, besides ja (yes) and nee (no). Based on the available polarity particles in a language, what polarity particle should be used to respond affirmatively and negatively to a negative polar question? The anaphor-based account (Krifka, 2013), the feature-anaphor-based account (Roelofsen & Farkas, 2015) and the constraint-based account (Hoek & de Hoop, 2016) all provide slightly different answers. Moreover, there is a lack of empirical data concerning responses to negative polar questions. How are ja, nee and
jawel used and interpreted in response to negative polar questions in Dutch?
The first study in this thesis investigated how Dutch adults respond to negative polar questions in Dutch. The results of an online questionnaire indicate that Dutch adults use nee to respond
affirmatively to a negative polar question and prefer to use jawel to respond negatively to a negative polar question, however ja and nee can also be used. The results of the questionnaire further indicated that Dutch adults prefer to add information to the bare polarity particle question when answering a negative polar question (e.g., yes, he did). What motivates Dutch adults to add information to the bare polarity particle? The possible ambiguity in interpreting a bare particle response to a negative polar question might be able to explain this phenomenon (Krifka, 2013; Roelofsen and Farkas, 2015).
Therefore, the second study investigated how answers to negative polar questions are interpreted in Dutch. The results of an online questionnaire indicated that there is no significant difference in interpretation between bare nee responses and nee responses with an expanded reply. They are both interpreted as affirmative responses to negative polar questions. Thus, there seems to be a discrepancy between production and interpretation. Participants tend to use an expanded reply in answering negative questions affirmatively, however, there is no difference in the interpretation of nee and nee with a negative expanded reply. The use of an expanded reply can therefore not be motivated by ambiguity in interpreting bare particle responses to negative polar questions.
Table of Contents
Introduction ...6
1.Theoretical Background ... 10
1.1 Anaphor-Based Account (Krifka, 2013) ... 10
1.1.1 Optimal Choice ... 11
1.1.3 The German Polarity Particle Doch ... 14
1.1.4 Extending Krifka’s Predictions to Dutch ... 16
1.2 Feature-Based Account (Roelofsen & Farkas, 2015) ... 17
1.2.1 Markedness ... 18
1.2.2 Feature Combinations ... 19
1.2.3 Doch ... 20
1.2.4 Extending Roelofsen and Farkas’ (2015) Prediction to Dutch ... 20
1.3 Constraint-Based Account (Hoek & de Hoop, 2016) ... 21
1.3.1 Prediction... 24
1.4 Empirical Studies ... 25
1.4.1. Acceptability and Interpretation of Bare Particle Answers to Negative Polar Questions ... 25
1.4.2 Types of Answers to Various Polar Questions... 26
1.4.3 Short Replies to Negative Polar Questions in Italian ... 27
1.5 Statement of Purpose and Hypothesis ... 29
1.6 Hypothesis ... 30
2. Experiment 1: Production of Polarity Particles ... 33
2.1. Method ... 33
2.1.2 Materials and Design ... 33
2.1.3 Procedure... 36
2.1.4 Data and Analysis ... 37
2.2 Results... 38
2.2.1 Responding Affirmatively to a Positive Polar Question ... 38
2.2.2 Responding Negatively to a Positive Polar Question ... 39
2.2.3 Responding Affirmatively to a Negative Polar Question ... 40
2.2.4 Responding Negatively to a Negative Polar Question... 42
2.3 Discussion... 43
2.3.1 Positive Polar Questions ... 43
2.3.2 Affirmative Responses to Negative Polar Questions... 44
2.3.3 Negative Responses to Negative Polar Questions ... 45
3. Experiment 2: Interpretation of Polarity Particles... 47
3.1. Method ... 47
3.1.1 Participants ... 47
3.1.2 Materials and Design ... 47
3.1.3 Procedure... 50
3.1.4 Data and Analysis ... 50
3.2 Results... 50
3.2.2 Positive Polar Questions ... 50
3.2.3 Negative Polar Questions... 51
3.3 Discussion... 52
4.1 Production... 55
4.1.1 Affirmative Responses ... 55
4.1.2 Negative Responses ... 57
4.2 Interpretation ... 57
4.3 Production and Perception ... 59
4.4 Relating Results to Empirical Studies ... 60
4.5 Revisiting Linguistic Theories ... 61
4.6 Limitations and Further Research... 63
5. Conclusion ... 64
6. References... 65
Production and Interpretation of Answers to Negative Polar Questions in Dutch
‘Don’t you want coffee?’, ‘Isn’t your mom coming?’ or ‘Haven’t you seen my keys?’. These are all negative polar questions: questions containing a negation that ask the listener to affirm or negate the proposition expressed by the question. Polarity particles such as yes and no can be used to respond to negative as well as positive polar questions. To answer positive polar questions it is clear what particle should be used. ‘Is your mom coming?’ ‘Yes, she is’ or ‘No, she isn’t’. Responding to negative polar questions such as ‘Isn’t your mom coming?’, is more complicated. You could both say ‘Yes, my mom is not coming’, and ‘No, my mom is not coming’, to affirm that your mom is not coming. Moreover, you could say ‘Yes, my mom is coming’, or ‘No, my mom is coming’ to negate that your mom is not coming, and state that she is in fact coming. So the question arises: what do yes and no actually mean when they are used as an answer to a negative polar question?
According to Pope (1973), two system can be distinguished in deciding how to answer a negative polar question. There are polarity-based systems (in languages such as English and Dutch) and truth-based systems (such as Japanese). A truth-based system states that a polarity particle affirms or negates the truth of the negative proposition in the question. In a polarity-based system, in contrast, the polarity particle agrees with the polarity of the (elided) proposition in the response (Holmberg, 2013). This means that based on a truth-based system such as Japanese, polarity particles are used as in the following example:
(1) Is your mom not coming? (negative proposition: Your mom is not coming) (a) No (= she is not not coming = she is coming)
(b) Yes (= she is not coming)
In contrast, in a polarity-based system such as English, the polarity particle mirrors the polarity of the response. Therefore, the following responses are predicted:
(2) Is your mom not coming? (a) Yes (= she is coming) (b) No (=she is not coming)
Based on Pope (1973) a language uses either a truth-based system or a polarity-based system. However, in reality it is possible that a language expresses features from both systems (Holmberg, 2013). In general each language has a preference to use a specific system in a specific situation. However, it is a simplified view to see English, for example, as a language that is either truth-based or polarity-based (Holmberg, 2013).
Moreover, not every language has the same polarity particles. While all languages have a version of yes and no, there are also language that have additional polarity particles. Romanian, for example, has a polarity particle (da) that needs to be included to negate the (salient) proposition of a question (Roelofsen & Farkas, 2015). Other languages might have a specific particle to negatively respond to a negative question or assertion. Following the terminology of Hoek and de Hoop (2016) a response to a polar question can be either affirmative or negative. To negatively respond to a polar question does not mean that a negative polarity particle such as no or nee should be used. A negative response indicates that the proposition of the question is negated by the polarity particle. German, for instance, has the word doch besides ja and nein (yes and no) that can be used to negatively respond to a negative polar questions. In Dutch the word jawel can be used to do the same:
(3) Wil je geen taart? Don’t you want cake? Jawel (, ik wil wel taart) Yes (, I do want cake).
Based on the different available polarity particles in languages, the question arises: what responses do speakers provide, given the existing polarity particles in a language, to negative polar questions? Speakers could use bare particle response or responses with an expanded reply. Expanded replies consist of additional linguistic material that add more details, an explanation or repeat part of the question in order to convey the intended polarity more explicitly. What type of response are used to respond to negative polar questions? Bare particle responses or responses with an expanded reply?
In (4) a model of a question-answer pattern is presented.
When applied to negative polar questions, the antecedent clause is a negative polar question (e.g., Don’t you want cake?). In response to the antecedent a polarity particle is used (e.g., yes), with an optional expanded reply (e.g., I do want cake). Krifka (2013) and Roelofsen and Farkas (2015) assume that expanded replies are used to avoid ambiguity in the interpretation of responses to negative polar question. This raises another question: how are polarity particles interpreted?
There are several accounts that try to explain what polarity particle should be used to respond to negative polar questions and how these particles are interpreted when used to respond to a negative polar question. In this thesis we will focus on the following three theories; the anaphor-based account (Krifka, 2013), the feature-based account (Roelofsen & Farkas, 2015), and the constraint-based account (Hoek & de Hoop, 2016). All three accounts focus on what answers are expected to negative polar questions for English. Krifka (2013) also predicts what responses are expected in German where
doch, a polarity particle that is used to negatively respond to a negative polar question, is available.
Roelofsen and Farkas (2015) try to account for all languages, and give specific examples for several languages including Romanian and German. Hoek and de Hoop (2016) give a detailed prediction of Dutch (see section 1 for a more detailed description of the three account).
The predictions of the three main theories differ slightly for languages that include a dedicated particle to negatively respond to a negative polar question such as German or Dutch. Moreover, there is a lack of empirical evidence from these languages. Therefore, it is relevant to investigate whether the empirical predictions of the three accounts hold. This study intends to provide empirical evidence for the claims made by the three accounts. It does so by investigating what polarity particles are used to respond to negative polar questions in a language that includes a specific particle to negatively respond to a negative polar question, namely Dutch. Dutch has three available polarity particles: ja,
nee, and jawel. One of the goals of this thesis is, therefore, to find an answer to the question: ‘How do
adult speakers of Dutch affirmatively and negatively respond to negative polar questions?’ This will be achieved by conducting an online experiment in which participants have to read short descriptions of a hypothetical situation and are asked to either affirm or negate the proposition in the question.
Additionally, since it is unclear how production of polarity particles is related to the interpretation of these particles, a second research question is: ‘How are answers to negative polar questions in Dutch interpreted?’. Do production and interpretation match? If a speaker replies no to the question ‘Don’t you want a cookie?’, does this mean that the speaker wants a cookie, or not? This question will be investigated through a second online experiment in which participants read short descriptions of hypothetical situations. The participants is asked how (s)he interprets a given answer to a polar question by choosing from two contradicting options. The hypothetical situations in the second experiment are identical to the hypothetical situations in experiment one.
First, however, a detailed description will be given of the three main theories that explain and predict how polarity particles are used, and how they are interpreted when used as response to a negative polar question. In addition, the results of the few empirical studies conducted to study answers to negative polar questions will be discussed. In the second part of this thesis the design and results of the two experiments will be discussed. Finally, the results of both studies will be compared and discussed.
1.Theoretical Background
In this section three main theories concerning answers to negative polar questions will be discussed: the anaphor-based account (Krifka, 2013), the feature-based account (Roelofsen & Farkas, 2015) and the constraint-based account (Hoek & de Hoop, 2016).
1.1 Anaphor-Based Account (Krifka, 2013)
Krifka (2013) proposes an account for the fact that while the production and interpretation of polarity particles (referred to as ‘response particles’ by Krifka) such as yes and no seems relatively straightforward, in fact, is more complex than one might think. Responding to an assertion or question containing a negation is particularly complicated. Krifka claims that the semantics of polarity particles are simple, however, it becomes more complex when one has to respond to an antecedent containing a negation. He accounts for this additional complexity by claiming that polarity particles such as yes and
no are propositional anaphors that pick up discourse referents, propositions under discussion, that are
salient. For example this explicit proposition under discussion:
(5) Did Ede steal the cookies?
Salient proposition: Ede stole the cookies
Yes = anaphor that picks up ‘Ede stole the cookies’
Contrary to positive polar questions, negative polar questions introduce two possible discourse
referents that can be picked up by a polarity particle. To motivate this assumption, Krifka (2013) gives the following example:
(6) Two plus two isn’t five. 1. Everyone knows that. [d’]
2. That would be a contradiction. [d]
In the example the first ‘that’ refers to the negative discourse referent [d’] = ¬ (2+2=5), while in the second sentence ‘that’ refers to positive discourse referent [d] = (2+2 = 5). A negative polar question therefore introduces both a positive and a negative discourse referent:
(7) Did Ede not steal the cookie? Propositional discourse referents:
d: Ede did not steal the cookie
d’: ¬ (Ede did not steal the cookie) = Ede stole the cookie
A polarity particle picks up only one of these discourse referents as its antecedent. In English there are two polarity particles: yes and no. Krifka (2013) states that yes picks up a salient
propositional discourse referent and asserts it, whereas no picks up salient discourse referent and asserts its negation. Since there are two salient discourse referents in a negative polar question, both
yes and no can pick up [d] (= Ede did not steal the cookie) and [d’] (= Ede stole the cookie), which
causes ambiguity. Responding with yes can both mean that Ede did and did not steal the cookie, depending on what discourse referent it picks up. The same is true for no. This results in the following possible form-meaning pairs when responding to a negative polar question such as in (7):
Table 1
Possible Form-Meaning Pairs in Responding to a Negative Polar Question According to Krifka (2013)
Possible Response to negative polar question Reference Resulting meaning
Yes d He didn’t
Yes d’ He did
No d He did
No d’ He didn’t
Based on the possible form-meaning pairs in Table 1, what is the optimal way to respond affirmatively and negatively to a negative polar question? The next section will explain how Krifka (2013) decides what the optimal form-meaning pair is.
1.1.1 Optimal Choice
To decide what the optimal choice is, Krifka (2013) uses Optimality Theory (OT). According to his OT account, in English there are two forms: yes and no that can express two different meanings:
Ede did steal the cookie and Ede did not steal the cookie. The optimal form and meaning are
determined by a set of constraints that differ in rank and can be violated. Linguistic rules are not hard according to OT, but violable. If a conflict arises between two constraints, it is important that the
highest ranked constraint is satisfied. A form-meaning pair can therefore violate a lower ranked constraint and still be the optimal choice (Prince & Smolensky, 1993/2004; Hendriks et al., 2010).
Krifka (2013) introduces two constraints which he argues to be relevant in determining the optimal choice to answer a negative polar question:
1 *NegDR: Penalizes picking up a negatively marked discourse referent. 2 *DisAgr: Penalizes disagreement with the other speaker.
These constraints are the result of two markedness principles. First of all, disagreement with the person who asks the question is marked. Therefore, *DisAgr is violated when a negative response is given to a negative polar question. Second, the two propositional discourse referents that are introduced by a negative polar question differ in saliency. To negate something, the non-negated proposition must already be salient within the context. Therefore, Krifka (2013) assumes that the non-negated proposition is more salient than the non-negated proposition. It is easier for an anaphor to pick up a salient referent as its antecedent than a non-salient referent and therefore the *NegDR constraint is violated when a negated proposition is picked up.
According to OT these constraints can differ in rank. Krifka (2013) claims that for negative assertions one can “plausibly assume” (p.13) that *DisAgr is ranked higher than *NegDr. For negative polar questions he states:
“the reactions are similar as to negated assertions, except that yes is more naturally interpreted as ‘he did’ (cf. Holmberg 2012). This would follow from the plausible assumption that questions based on negated propositions do express a bias, but a weaker one than assertions. Hence the constraint *DisAgr is ranked lower [than *NegDr] […]” (Krifka, 2013, p.13)
*NegDr is thus ranked higher than DisAgr* for negative polar questions, since for questions, in contrast to assertions, there is only a weak bias towards one of the two discourse referents (Krifka, 2013). It must be noted that using a different ranking of constraints for different linguistic
This constraint interaction is represented in the following OT tableau (Figure 1). The input is a polarity particle in combination with a the salient proposition. This combination is the ‘form’ of the form-meaning pair. The output is the resulting meaning. For instance, yes can pick up [d] (= Ede did not steal the cookie) resulting in the meaning that Ede didn’t steal a cookie. An asterisk in the column headed by a particular constraint means that the combination of input and output in the corresponding row violates that constraint. If a constraint is violated this candidate is less optimal than an option for which no constraints are violated. Sometimes, however, there is no optimal form-meaning pair possible without any constraint being violated. In that case the rank becomes important. If one of the form-meaning pairs violates a lower ranked constraint, this is a better option than a form-meaning pair that violates a higher ranked constraint. The little hand with the pointed finger expresses the optimal outcome. The pointed finger in brackets expresses the second best outcome, and the pointed finger between two brackets the third best outcome.
Figure 1
Optimality Theory Tableau for English Polarity Particles, Antecedent: ‘Did Ede not steal the cookie?’
Input Output Constraints
Polarity Particle Reference Resulting meaning *NegDr *DisAgr
(()) A Yes D He didn’t *
() B Yes d’ He did *
C No D He did * *
D No d’ He didn’t
According to the tableau in Figure 1, the optimal candidate is (D) no meaning ‘he didn’t’, because this does not violate any constraint. Therefore, no is predicted to be understood as an
affirmative response to a negative polar question. According to OT, this is the only optimal candidate. Krifka (2013), however, also selects a second best (B) and third best (A) option. The preferred interpretation of yes is therefore as a negative response to a negative polar question (B). (B) violates *DisAgr, however since it has a lower ranking than *NegDr that is violated by (A) it is preferred over (A). Since the difference between the expressed meaning of yes in (A) and (B) only differs slightly,
Krifka (2013) states that usually an expanded reply (in his terminology: elliptical clause) is required to express the correct meaning, to avoid ambiguity. (C) is not an option according to Figure 1, however Krifka (2013) states that no can still be interpreted as a negative response to a negative question (option C), but in that case it should be followed by an expanded reply, resulting in: ‘no, he did’.
Based on Krifka’s theory one would expect that when a speaker affirmatively answers a negative polar question (s)he is predicted to answer no, whereas when (s)he wants to respond negatively to a negative polar question (s)he is predicted to answer yes (Krifka, 2013). However, he does not completely eliminate the use of no as a negative response and yes as an affirmative response to a negative polar question.
1.1.3 The German Polarity Particle Doch
In the previous section Krifka’s (2013) predictions for a language with just two polarity particles yes and no, like English, are described. However, what about answers to negative polar questions in German, which has another polarity particle: doch that is specifically used to negatively respond to negative polar questions (and assertions)? Two new constraint are introduced:
3 *Pres: doch presupposes two discourse referents to be salient, and picks up the non-negated discourse referent and asserts it.
4 Block: if there is an expression present for which the indicated interpretation is strongly preferred, that expression should be used.
Krifka (2013) states that doch presupposes that at least two discourse referents are salient. Positive questions and assertions only make salient one discourse referent. Therefore, doch can solely be used to respond to negated assertions and questions. Moreover, doch picks up a non-negated discourse referent and asserts it. This is expressed by the constraint *Pres (short for presupposition). *Pres is violated when doch asserts the negated discourse referent. The German particle doch is therefore in competition with ja (German yes) and nein (German no). It seems that the option of ja to pick up [d’] (= Ede stole the cookie) is blocked by the existence of doch. This is expressed by the constraint Block (Beaver, 2004). Block is violated if the ja picks up [d’] and is used to negatively
respond to a negative question, because for this interpretation the use of jawel is strongly preferred. It is unclear why nee is not blocked by doch.
Figure 2
Optimality Theory Tableau for German Polarity Particles, Antecedent: ‘Did Ede not steal the cookie?’
Input Output Constraints
expression reference resulting meaning *Pres Block *NegDR
() A ja d He didn’t * B ja d’ He did * () C nein d He did * D nein d’ He didn’t E doch d He didn’t * * F doch d’ He did
In Figure 2 DisAgr* is irrelevant under meta-constraint Block (Krifka, 2013). DisAgr* is violated when there is disagreement with the other speaker. This is the case for doch, which would result in it not being the optimal choice, while it is. By introducing Block the disagreement in (B) is violated by Block, and the disagreement with the speaker in C is violated by *NegDr, therefore making DisAgr* irrelevant. Moreover, by introducing the meta-constraint Block, Krifka integrates production into an interpretation tableau. If this tableau were to focus solely on interpretation the outcome would be independent of another form-meaning pair being present for which a certain interpretation is strongly preferred. Similar to Figure 1, in Figure 2 Krifka (2013) not only selects optimal candidates (D & F) but also second best candidates (A & C), which is not conventional in OT.
In Figure 2, the presence of doch blocks the interpretation of ja as he did (B). This tableau predicts that doch will be interpreted as a negative response to a negative polar question (F) and is an optimal candidate. Nein will be interpreted as an affirmative response to a negative polar question (D) and is the second optimal candidate. Moreover, nein (C) is not blocked by doch, but dispreferred, since it is a second best option. Ja can also be interpreted as a negative response to a negative polar question
(A), however, this is dispreferred (Krifka, 2013). In contrast to Figure 1, there is no third best option in Figure 2 and option (B) and (E) are not regarded as possible interpretations.
1.1.4 Extending Krifka’s Predictions to Dutch
The research question in this paper focuses on Dutch. While Krifka (2013) does not explicitly makes any prediction about Dutch, Dutch has a polarity particle similar to the polarity particle doch in German: jawel. Jawel is (in the same way as doch) used to negatively respond to a negative polar question. Based on Krifka’s (2013) theory and analysis of German, the prediction for production of answers to negative polar questions in Dutch is as set out in Table 2. When Krifka’s (2013) prediction for German is extended to Dutch this means that when answering a negative polar question
affirmatively nee and ja can be used, however, ja is dispreferred. To answer a negative polar question negatively jawel and nee can be used, however, nee is dispreferred.
Table 2
Krifka’s (2013) Prediction of the Production of Polarity Particles as Response to Polar Questions
Question Response Production
negative affirmative nee (or ja)
negative negative jawel (or nee)
In Table 3 Krifka’s (2013) prediction of the interpretation of polarity particles as response to negative polar questions is presented. Since nee can be used to both affirmatively and negatively respond to a negative polar question, but the interpretation of nee as an affirmative response is preferred, an exanded reply is needed to use nee as a negative response to a negative polar question.
Table 3
Krifka’s (2013) Prediction of the Interpretation of Polarity Particles as Response to Negative Polar Questions
Interpretation ja affirmative
nee affirmative (or negative)
jawel negative
1.2 Feature-Based Account (Roelofsen & Farkas, 2015)
The aim of the study of Roelofsen and Farkas (2015) is to create a theory of production and interpretation of answers to polar question in all languages. They try to achieve this by listing all the phenomena their theory should be able to account for. The empirical results of the study of Kramer and Rawlins (2012) (see section 1.4.1) is one of these phenomena. Roelofsen and Farkas (2015) conclude that polarity particles are anaphoric and are not anaphors themselves as Krifka (2013) states. There is an anaphoric link (8) between the antecedent and the (possibly elided) prejacent (Roelofsen & Farkas, 2015). ‘Prejacent’ is term borrowed from logic and refers to a proposition that is laid out previously (in the antecedent).
(8) (a) [Antecedent clause]…..[yes/no[Prejacent]] Anaphoric link
(b) [Don’t you want coffee? ]…….[yes/no[I (don’t)want coffee]]
Anaphoric link
Roelofsen and Farkas (2015) assume that polarity particles are words that function as the morphological realization of two types of polarity features: absolute and relative features. The absolute polarity features indicate whether a response to a polar question is either a positive [+] or a negative [-] proposition. A relative polarity feature marks the relation between a response and the antecedent as either agreeing; [agree], or reversing; [reverse]. Each polarity particle both realizes an absolute and relative feature and therefore has a double ‘function’. This results in four possible combinations:
(9) Relative Absolute
[agree, +] agree +
[agree, −] agree −
[reverse, +] reverse +
[reverse, −] reverse −
[agree, +] is an affirming response to a positive polar question, [agree,-] an affirming response to a negative polar question, [reverse, +] a negative response to a negative question, and [reverse, -] a negative response to a positive question.
1.2.1 Markedness
Based on the possible combinations of features in (9) a polarity particle has to be chosen by the speaker to convey a certain meaning. The English polarity particles yes and no both realize an absolute and relative feature simultaneously. However, not all combination of features fit optimally with the features that are realized by the polarity particles. In English [agree] and [+] are realized by
yes and [reverse] and [-] are realized by no. The choice of a polarity particle for the realization of
[agree, +] and [reverse, -] is therefore simple. Yes realizes [agree, +] and no realizes [reverse, -]. However, what about [agree, -] and [reverse, +]? Yes and no can only realize one of the two features in these combinations. Yes should be used to realize [agree] in [agree, -], however, no should be used to realize [-]. And no should be used to realize [reverse] for the feature combination [reverse, +], however yes should be used to realize [+]. Only one of the two features can be realized at once. How to decide which of the two feature should be realized?
Roelofsen and Farkas (2015) assume that features differ in markedness based on the
observations of Pope (1976). The more marked a feature is, the stronger the need for it to be realized. [+] is less marked than [-] and [agree] is less marked than [reverse]. This means that when a polarity particle has to be selected for [reverse,+], [reverse] has higher realization needs than [+] and therefore
no, which realizes [reverse] would be the optimal candidate. However, Roelofsen and Farkas (2015)
[reverse] responses are marked because they contrast with the polarity of their antecedent” (p.388). This means that in [reverse, +] responses both [reverse] and [+] have high realization needs. [reverse] because it is more marked than [agree], and [+] because it stands in contrast with [reverse]. Instead of
no as the best candidate to realize the highest realization needs, both yes and no are equally
(sub)optimal. The overall markedness scale from least marked to most marked is:
(10) [agree, +] < [reverse, −] < [agree, −] < [reverse, +]
Based on the markedness scale and the realization needs of the different features it can be decided what polarity particle should be used to affirmatively and negatively respond to a negative polar question (Roelofsen & Farkas, 2015).
1.2.2 Feature Combinations
As stated above in (9) four different feature combinations are possible. Roelofsen and Farkas (2015) predict what polarity particle is used for these feature combinations. They argue that yes is used to realize [agree, +] (i.e., 11a below) and no is used to realize [reverse, −] (i.e., 11b below). What about [agree, -] (i.e., 11c)? According to Roelofsen and Farkas (2015) this can be realized by both yes and no, since yes realizes [agree] and no realizes [-]. [-], however, is a marked feature and therefore has higher realization needs. Therefore a preference for no is expected. In [reverse, +] (11d below) both [reverse] and [+] are marked ([+] is marked because it stands in contrast with [reverse] according to the contrastive markedness principle) so both yes and no are possible answers to negatively respond to a negative polar question:
(11) a. [agree, +] can only be realized by yes. Did Peter call? Yes (he did) b. [reverse, −] can only be realized by no. Did Peter call? No (he didn’t)
c. [agree, −] can be realized by no or yes, Didn’t Peter call? Yes/no (he didn’t) no is preferred.
d. [reverse, +] can be realized by yes or no, Didn’t Peter call? Yes/no (he did) no preference is expected.
Bare particle responses to negative polar questions are predicted to be ambiguous (Roelofsen & Farkas, 2015). Both yes and no can be interpreted as an [agree, −] response, an affirmative response to a negative polar question, or as [reverse, +], a negative response to a negative polar question. Moreover, Roelofsen and Farkas (2015) state that to affirm a negative polar question an expanded reply (in their terminology: prejacent) can be used to disambiguate the answer. To negatively respond to a negative polar question in English, however, there is additional pressure to add an explicit expanded reply (e.g. ‘yes, he did’ or ‘no, he did’). This is because there is an independent reason why a [reverse,+] answer needs an expanded reply, namely to stress the contrast between [reverse] and [+]1. If no prejacent is added, the intended [reverse, +] response will most likely be interpreted as an [agree, -] response (Roelofsen & Farkas, 2015).
1.2.3 Doch
English only has two polarity particles yes and no. German, however, has an extra polarity particle doch, a specific polarity particle used to negatively reply to a negative polar question. What about the answering pattern to negative polar questions in German? According to Roelofsen and Farkas (2015) the two German polarity particles, ja and nein, are assumed to have the same realization potential as English yes and no. This means that [agree] and [+] can be realized by ja and [reverse] and [-] can be realized by nein. However, they state that both ja and nein are blocked in [reverse, +] responses because there a dedicated [reverse, +] particle, doch available. How and why jawel blocks the use of ja and nee is not discussed by the authors.
1.2.4 Extending Roelofsen and Farkas’ (2015) Prediction to Dutch
The present study focuses on Dutch. However, Roelofsen and Farkas (2015) do not make any explicit predictions for Dutch. Since the German particle doch and the Dutch particle jawel behave similarly, and are both used to negatively respond to a negative polar question, their prediction for German could be extended to Dutch. Based on the analysis of Roelofsen and Farkas (2015) for
1 It is unclea r from the feature-based account (Roelofsen & Fa rkas, 2015) why this does not a pply to [a gree, -],
German, the following predictions can be formulated for the production (Table 4) and interpretation (Table 5) of answers to negative polar questions in Dutch.
Table 4
Roelofsen and Farkas’(2015) Prediction of the Production of Polarity Particles as Response to Polar Questions
Question Response Production
positive affirmative Ja
positive negative Nee
negative affirmative nee (preferred) or ja
negative negative jawel
Table 5
Roelofsen and Farkas’(2015) Prediction of the Interpretation of Polarity Particles as Response to Negative Polar Questions
Interpretation
ja affirmative or negative
nee affirmative or negative
jawel negative
It must be noted that there seems to be a discrepancy in the theory of Roelofsen and Farkas (2015) between production and interpretation. While nee and ja can only be used to affirmatively respond to a negative polar question, they can be interpreted as both an affirmative and a negative response to negative polar question, because of the features they realize. If the speaker takes this ambiguity into account, (s)he might use an expanded reply.
1.3 Constraint-Based Account (Hoek & de Hoop, 2016)
Hoek and de Hoop (2016) looked into the various interpretations of ja and nee in Dutch. They use bi-directional Optimality theory to explain how speakers select the correct polarity particle to convey a certain meaning and listeners select the correct interpretation of these polarity particles,
given the context. Krifka (2013) used unidirectional Optimality Theory to find the optimal form-meaning pair to either affirmatively of negatively respond to a negative polar question. Hoek and de Hoop (2016) use bi-directional optimality theory. This means that their theory combines the
production perspective with the comprehension perspective, instead of only using the interpretation perspective, like Krifka (2013). It accounts for the intuition that the listener has to consider the speaker’s perspective, and the speaker has to consider the listener’s perspective (Hendriks et al., 2010).
According to Hoek and de Hoop (2016) there are four constraints at work:
1 ALIGN: When answering affirmatively the polarity of the polarity particle and the question should align. When answering negatively the polarity of the polarity particle and the polarity of the question should dis-align.
2 FAITH: use ja to give an affirmative response and nee to give a negative response . 3 *NEG: avoid negation in the output.
4 ECONOMY: is violated when other forms than ja or nee are used.
ALIGN is ranked higher than FAITH, because otherwise ja would be used to affirmatively answer all polar questions and nee would be used to negatively respond to all polar questions. For positive questions these two constraints give a similar result, however, they conflict when answering negative questions. The ALIGN and FAITH constraint are based on the features of Roelofsen & Farkas (2015). The FAITH constraint represents the absolute polarity ([+] and [-]) and the ALIGN constraint
represents relative polarity ([same] and [reverse]) (Hoek & de Hoop, 2016). *NEG captures two different principles and can therefore be violated twice. Firstly, the use of a negation (nee) violates the constraint because utterances containing a negation are more complex. Moreover, answers that negate the proposition of the question violate the constraint because speakers tend to avoid disagreement.
*NEG is ranked below FAITH because being faithful is more important than avoiding marked responses. Finally, ECONOMY is violated when a more complex form than just ja or nee is used to respond, for instance jawel. This constraint is also violated if an expanded reply would be used. This
constraint is ranked the lowest, because otherwise jawel would never be used. In Figure 3 the bidirectional OT Tableau with form-meaning pairs of Dutch polarity particles is presented.
Figure 3
Tableau in Bidirectional OT for Answers to Polar Questions in Dutch (cf. Hoek & De Hoop)
Form-meaning pairs Constraints
Form Meaning ALIGN FAITH *NEG ECONOMY
A ja aff + B nee aff + * * * C jawel aff + * D ja neg + * * * E nee neg + ** F jawel neg + * * * * G ja aff - * H nee aff - * * I jawel aff - * * J ja neg - * * K nee neg - * ** L jawel neg - * * *
In Figure 3 ALIGN is violated when nee is used to affirm a positive question, and when ja or
jawel are used to affirm a negative polar question. Moreover, this constraint is violated when ja or jawel are used to negate a positive question and when nee is used to negatively respond to a negative
polar question. FAITH is violated when ja is used to negatively respond, nee is used to affirmatively respond, or jawel is used. *Neg is violated when the response is negative, or nee is used. Finally, ECONOMY is violated by jawel.
The bi-directional optimal candidate is A, since it does not violate any constraint. Because ja is the optimal candidate to respond positively to a positive utterance, B and C are blocked, since these forms convey the same meaning as A. D, G and J are blocked because they include the same form. A second round of optimization reveals that E, responding negatively to a positive response using nee, is the next best bi-directional candidate because it does not violate ALIGN and FAITH. This blocks F,
because this candidate has the same meaning as E, and H and K because these candidates have the same form as F. A third round of optimization reveals that jawel is the optimal candidate to respond negatively to a negative utterance. This blocks I because candidate I has the same form. No form-meaning pairs are left. However, Hoek and de Hoop (2016) proceed from the point of view that there are four possible meanings, that all have to be expressed using a suitable form, even though all possible options are blocked. Therefore they use unidirectional OT to derive that to respond affirmatively to a negative utterance nee is the optimal form (Hoek & de Hoop, 2016). It must be noted that switching from bi-directional OT to unidirectional OT is a highly unconventional move in OT.
1.3.1 Prediction
Hoek and de Hoop (2016) use bi-directional OT to find the optimal form-meaning pairs for polarity particles in Dutch. The predictions of Hoek and de Hoop (2016) for production of answers to polar questions are presented in Table 6.
Table 6
Hoek and de Hoop’s (2016) Prediction of the Production of Polarity Particles as Response to Polar Questions
Question Response Production
positive affirmative ja
positive negative nee
negative affirmative nee
negative negative jawel
The prediction of the interpretation of answers to negative polar questions in Dutch of Hoek and de Hoop (2016) can be found in Table 7. Hoek and de Hoop (2016) do not expect ja to be used as a response to a negative polar question, their account therefore only predicts an interpretation of ja as response to a positive question, and not to a negative polar question.
Table 7
Hoek and de Hoop’s (2016) Prediction of the Interpretation of Polarity Particles as Response to Negative Polar Questions
Interpretation
ja n.a.
nee affirmative
jawel negative
1.4 Empirical Studies
In this section the empirical studies concerning answers to negative polar questions will be discussed. Thus far, previous studies have mainly focussed on the acceptance rate of responses to negative assertions (Brasoveanu, Farkas, & Roelofsen, 2013; Claus, Meijer, Repp, & Krifka, 2017; Meijer, Claus, Repp, & Krifka, 2015) instead of questions. There is a lack of empirical studies investigating responses to negative polar questions. Most research on responses to negative polar questions is purely theoretical and focusses on the syntactic and semantic characteristics of negative polar questions (Krifka, 2013; Hoek & de Hoop, 2016; Roelofsen & Farkas, 2015). However, a small number of studies empirically investigated the subject. These studies will be briefly discussed below.
1.4.1. Acceptability and Interpretation of Bare Particle Answers to Negative Polar Questions Roelofsen and Farkas (2015) used the results of the empirical study of Kramer and Rawlins (2012) as a guiding principle in developing their theory of the use and interpretation of polarity particles. Kramer and Rawlins (2012) investigated the acceptance of bare particle responses to negative polar questions. They used an acceptability judgement task in which participants had to indicate the felicity of a given answer on a seven point scale. Moreover, participants had to assess the truth of the answer, given the context, by choosing either ‘true’, ‘false’ or ‘unsure’. Kramer and Rawlins (2012) used items such as (12) that each consisted of a short written context (either positive or negative), a question (either positive or negative) and a bare particle response (yes or no).
(12) Context: John was planning to take a vacation soon. Bill is John’s boss and knows that he didn’t go on vacation. Sue just noticed John’s car in the parking lot.] (negative context)
Sue: Did John not take his vacation? (negative polar question) Bill: Yes. (positive response)
The results of the acceptability judgement task indicated that a bare yes or no response to a negative polar question is rated as less felicitous than a ‘true’ rated yes or no to a positive polar question. Furthermore, the results suggest that when answering a negative polar question when the context is positive, a bare yes and a bare no are less felicitous compared to when the context is negative (Kramer & Rawlins, 2012). Truth value judgements showed that bare particle responses to negative polar questions are ambiguous. Moreover, both bare yes and no are interpreted more often as an affirming response than as a negative response to a negative polar question (Kramer & Rawlins 2012).
1.4.2 Types of Answers to Various Polar Questions
Another study that empirically investigated responses to polar question is Enfield et al. (2019). The authors explored how polar questions are answered in fourteen different languages, including Dutch. The main goal of their study was to compare two strategies in answering polar questions across 14 different languages. Do language-users tend to use interjection-type answers, which include bare particle answers (yes, no, uh-huh, etc.) or do they prefer repetition-type answers, which repeat a part or all words of a question without using a polarity particle? They distinguished three different types of questions: requests for information, requests for confirmation and initiations of repair. The results of the study suggest that all fourteen languages (including Dutch) use both interjection-type and
repetition-type answers. Moreover, there seems to be a strong preference for interjection-type answers
to respond to polar questions (Enfield et al., 2019).
Unfortunately no distinction was made between positive and negative polar questions. The data used in these experiments were videos of informal, spontaneous everyday data. Since negative polar questions are marked compared to positive questions (Hoek & de Hoop, 2016) they are usually
avoided in conversation. Therefore asking a negative polar question in spontaneous everyday speech is very uncommon. This means that while the results of Enfield et al., (2019) do give a valuable insight into what type of answers are given to polar questions, it does not provide any specific insight in answers to negative polar questions.
1.4.3 Short Replies to Negative Polar Questions in Italian
Another empirical study investigating responses to negative polar questions is the study of Andorno and Rosi (2015). They investigated short replies to negative polar questions in Italian. The aim of the study was to describe how a positive or negative polar question or assertion affects the response in terms of choice of polarity particle (sì and no) and other verbal strategies (e.g. using an expanded reply in order to disambiguate the answer). Based on a theory of preferential agreement from Conversation Analysis by Pomerantz (1984), it is expected that a negative response to a negative polar question or assertion is dispreferred compared to an affirming response. Moreover, negative responses to assertions are marked in comparison to negative responses to questions because of the difference in bias between questions and assertions (Andorno & Rosi, 2015).
Negative questions might seem unbiased, since they are questions. However, they can be either be positively or negatively biased. Negative questions are never asked to introduce a negative proposition but are used to reject propositions that are already at issue. They can be used to challenge the belief of the listener (rhetorical question) or to ask if (s)he affirms or negates a proposition raised previously (Andorno & Rosi, 2015). Reese (2007) states that when asking a positively biased negative question a speaker asserts a positive proposition (Isn’t there a vegetarian restaurant around here? Proposition: there is a vegetarian restaurant around here)) and asks the listener whether or not (s)he would reject it. When asking a negatively biased negative question, the speaker found some negative contextual evidence that clashes with this positive assumption (Isn’t there a vegetarian restaurant around here? Proposition: there is a vegetarian restaurant around here. Context: you only see a park, and some houses in the near distance) and ask for a confirmation. In their study Andorno and Rosi (2015) used negatively biased negative questions and assertions. Moreover, they distinguished how confident a speaker was towards the negative bias. Based on prosody they discern declarative and
interrogative sentences. When uttering a declarative sentence a speaker is more confident towards a negative bias, while an interrogative demonstrates that a speakers is less confident towards the negative bias (Andorno & Rosi, 2015).
The data was elicited in two different spoken tasks. The first task was an interview. In this interview the researcher asked the participants questions about their knowledge of several subjects, all related to foreign languages. In the second task (based on Anderson et al., 1991) the participant had a map with a specific route on it. The researcher had a slightly different map without the route. The goal of this task was for the participant to help the researcher draw the map. Both tasks lend themselves to ask negative polar questions to check certain assumption (Is X not situated left of X?). The actor who played researcher was explicitly instructed to use as many negative questions and assertions as possible to elicit as many answers to negatively framed utterances as possible (Andorno & Rosi, 2015).
The results suggest that no can be used to affirmatively respond to negative polar questions and assertions. In 61-74% of the cases (depending on the task) an expanded reply was used. To negatively respond to negative polar questions (and assertions), both sì and no can be used. However, they are almost always accompanied by an expanded reply to disambiguate the response (in 88-98% of all cases, depending on the task) (Andorno & Rosi, 2015). Furthermore, in the first task no was used more frequently to negatively respond to declarative sentences than to interrogative sentences. This is in line with the expectation that declaratives have stronger assumptions compared to interrogatives. The utterer of the reply wants to explicitly mark the rejection of these assumptions (Andorno & Rosi, 2015).
Only a few empirical studies have been conducted in the field of answers to negative polar questions. However, only one explored the production of answers to negative polar questions
(Andorno & Rosi, 2015) for Italian. Both in Italian and in English answering polar questions can result in ambiguity. Therefore, Italians often use an expanded reply in their production of polarity particles as response to a negative polar question. Moreover, Kramer and Rawlins (2012) report that for English, the interpretation of bare particle responses to negative polar questions results in ambiguity.
Dutch, compared to English and Italian, has an additional particle besides ja and nee to negatively respond to a negative polar question: jawel. How does this extra polarity particle affect the way negative polar questions are answered and interpreted?
1.5 Statement of Purpose and Hypothesis
In short, the anaphor-based account of Krifka (2013) states that polarity particles are anaphors that pick up discourse referents that are introduced by a negative polar question. Based on a
unidirectional OT account, Krifka (2013) predicts that to respond to a negative polar question negatively jawel and nee can be used; however, nee is dispreferred. To answer a negative polar question affirmatively nee and ja can be used; however, nee is preferred. This means that when interpreting nee as response to a negative polar question this particle could both be interpreted as an affirmative and a negative response to a negative polar question. Ja can only be interpreted as affirming a negative polar question, and jawel can be interpreted to negatively respond to a negative polar question.
The feature-based account of Roelofsen and Farkas (2015) states that there is an anaphoric link between a negative polar question and the (possibly elided) prejacent. The choice of polarity particle in answering a polar question depends on what features the particles are able to realize. The markedness of these features determines what polarity particle is used to answer a polar question. To answer a negative polar question affirmatively nee and ja can be used. However, a strong preference for nee is expected based on markedness. To answer a negative polar question negatively jawel is the best option. In interpreting answers to negative polar questions, this means that ja and nee can be interpreted both as an affirmative and a negative response to a negative polar question, since ja realizes [agree] and [+] and nee both [reverse] and [-]. Jawel can only be interpreted as a negative response to a negative polar question.
Finally, the constraint-based account of Hoek and de Hoop (2016) uses bidirectional Optimality Theory and combines production and interpretation of polarity particles. They expect nee to be used and interpreted as an affirmative response to negative polar questions and jawel as a
negative response to negative polar questions. In the current study both the production and interpretation of answers to negative polar questions will be investigated.
1.6 Hypothesis
The predictions of the three different accounts for production can be found in Table 8.
Table 8
Prediction of the Production of Polarity Particles as Response to Polar Questions by the Three Different Accounts.
Question Response Anaphor account
(Krifka)
Feature account (Roelofsen & Farkas)
Constraint account (Hoek & de Hoop)
positive affirmative - ja ja
positive negative - nee nee
negative affirmative nee (or ja) preferably nee Nee
negative negative jawel (or nee) jawel Jawel
Based on the predictions of the different accounts it is expected that to affirmatively answer a negative polar question, participants will use nee. According to Roelofsen and Farkas (2015) and Krifka (2013) it is also possible that participants affirm a negative polar question using ja. None of the accounts expect jawel to be an option. To negatively respond to a negative polar question it is
expected that participants will use jawel. According to Krifka (2013) nee is another option, however this is dispreferred. None of the accounts expect ja to be used to negatively respond to a negative polar question.
Moreover, since according to Krifka (2013) nee is ambiguous, but the preferred reading is an affirmative response, an expanded reply should be used when nee is used as a negative response to a negative question. Roelofsen and Farkas (2015) state that an expanded reply can be used to
disambiguate an answer, or stress the contrast between the polarity of the question and the response. While no ambiguity is expected in using a certain polarity particle to express a certain meaning, there is ambiguity in interpreting these particles as responses to negative polar question. Responses to negative polar question either have the form of [agree, -] or [reverse, +]. Yes realizes [agree] and [+]
and no realizes [reverse] and [-]. As response to a negative question both yes and no can be interpreted as either an affirmative or negative response. A speaker can anticipate on this possible ambiguity and might use an expanded reply. Moreover, the use of an expanded reply is expected in answering jawel in [reverse, +] to a negative polar question to stress the contrast between the polarity of the question and the response. However, since jawel is a dedicated particle to negatively reply to negative
questions, this might not be necessary. The authors do not explicitly comment on this. Finally, Hoek and de Hoop (2016) only state that for English an expanded reply is needed disambiguate a response to a negative polar question (Bald, 1980). However, since no ambiguity is expected in answering negative polar question in Dutch, and Hoek and the Hoop (2016) do not cover this subject, no explicit prediction can be derived from their account.
When interpreting bare particle answers (see Table 9) to negative polar questions it is expected that ja is interpreted as an affirmative (Krifka, 2013; Farkas & Roelofsen, 2015) or negative response (Farkas & Roelofsen, 2015) to a negative polar question. Hoek and de Hoop (2016) do not expect ja to be used as a response to a negative polar question. Nee is expected to be interpreted as an affirmative response (Hoek & de Hoop, 2016) or as either an affirmative or negative response (Krifka, 2013; Farkas & Roelofsen, 2015) a negative polar question. Finally, jawel is expected to be interpreted as a negative response to a negative polar question by all the accounts (Krifka, 2013; Farkas & Roelofsen, 2015; Hoek & de Hoop, 2016).
Table 9
Prediction of Interpretation of Polarity Particles as Response to Negative Polar Questions by the Different Accounts
Interpretation
Anaphor-based account Feature-based account Constraint-based account
ja affirmative affirmative or negative n.a.
nee affirmative (or negative) affirmative or negative affirmative
In the following sections the hypothesis put forward by the accounts are tested for Dutch. First, a production experiment will be presented, followed by an interpretation experiment. Finally, the results of the production and interpretation experiments will be compared in order to determine how production and perception relate to each other.
2. Experiment 1: Production of Polarity Particles 2.1. Method
2.1.1 Participants
In total 63 participants (14 male) participated in experiment 1. Based on exclusion criteria established prior to the experiment, participants were excluded if they were not a native speaker of Dutch (1 participant, 0 male), or if they were bilingual (in the instruction it was specified that one was a bilingual if (s)he learned a second language outside an educational setting before the age of 4) (6 participants, 2 male). Moreover, participants with cognitive disorders such as ADHD, Autism Spectrum Disorder (ASD) and dyslexia were excluded (6 participants, 2 male), participants who fell outside the age range of 18-30 (6 participants, 1 male) and participants who were not or had not been university students (5 participants, 0 male). This resulted in a group of 39 participants (9 male) with a mean age of 24 (age range: 20-30).
2.1.2 Materials and Design
To assess what answers participants give to negative polar questions, a questionnaire was made in Qualtrics (Qualtrics, 2005: Version: April 2020), an online survey software program. Participants read a short description of a particular situation after which a question was asked to them. They were asked to answer that question by typing in a response. It was indicated whether they had to respond affirmatively or negatively by a short sentence between brackets (see sample items (13) and (14)). They could type their response in a small text box in Qualtrics (Qualtrics, 2005). An example of a test item can be found in (13) and (14). Negations are easily overlooked, therefore, they were underlined to ensure participants will notice them.
(13) Lize is haar sleutel kwijt. Ze zoekt overal. Ze vraagt je: ‘Heb je mijn
sleutel niet gezien?’. (Je hebt de sleutel niet gezien).
Wat antwoord je?
English translation of (13):
(14) Lize has lost her key. She’s looking everywhere. She asks you: ‘Haven’t you seen my
How would you respond?
The experiment had a 2x2 design with factors polarity of the question (positive vs. negative) and polarity of the answer (affirmative vs. negative). The questionnaire consisted of 40 questions in total (for a complete overview of the items, see Appendix A). Sixteen of these items are test items and consisted of a negative polar question, of which eight included the word niet and eight included the word geen.. Eight items contained a positive polar question and were included as control items to see how participants respond to positive polar questions. Only eight control items were used, instead of the desired sixteen, because of time considerations. Another sixteen items were filler items. These filler items contained a disjunctive question. The filler items were used to mask the goal of the study. In each of the three item categories, one half of the items contained an affirmative target answer, and the other half of the items contained a negative target answer. Within each item category with a negative target answer, one half of the target answers included the word niet and the other half of the items included the word geen. Two versions of all test items, control items and filler item were made, one with an affirmative and one with a negative target answer. Based on these two versions, two different lists were made. If in the first list the item had a negative target answer, in the second list it had a positive target answer. The items within each list of the questionnaire were distributed as in figure 1. Qualtrics (Qualtrics, 2005) randomly assigned one of the lists to each participant. Within each list Qualtrics randomized the order of the items.
Figure 4
Distribution of Test Items, Control Items and Filler Items per Condition
To make sure the negative polar questions did not have a bias towards either a positive or a negative response in combination with the situation described, a pretest was conducted. In this pretest 5 participants were presented with a situation followed by a negative polar question (15 & 16) (for all items see Appendix B). Participants had to indicate how likely, given the situation, they thought a certain answer to a question was on a five point scale (very likely to very unlikely). For the actual test, only items were used where no more than 1 participant judged the answer to be unlikely.
(15) Lize is haar sleutel kwijt. Ze zoekt overal. Ze vraagt Tom: ‘Heb je mijn sleutel niet gezien?’.
Hoe waarschijnlijk vind je het dat Tom Lize haar sleutel heeft gezien? o heel waarschijnlijk
o waarschijnlijk
o even waarschijnlijk als onwaarschijnlijk o onwaarschijnlijk o heel onwaarschijnlijk English translation of (15): 40 items 16 test items negative polar questions 8x affirmative answer 4x niet 4x geen 8x negative answer 4x niet 4x geen 8 control items positive polar questions 4x affirmative answer 4x negative answer 2x niet 2x geen 16 filler items disjunctive questions 8x positive anwer 8x negative answer 4x niet 4x geen
(16) Lize has lost her key. She’s looking everywhere. She asks Tom: ‘Haven’t you seen my key?’
How likely do you think it is that Tom has seen Lize’s key? o very likely
o likely
o equally likely as unlikely o unlikely
o very unlikely
2.1.3 Procedure
All participants were recruited online via Facebook and Twitter. The message to recruit the participants can be found in Appendix C. Prior to the start of the experiment participants read a short introduction (Appendix D) after which they had to fill in an online informed consent form (Appendix E) and some general information (Appendix F) which was followed by the instructions of the test (Appendix G). In the instructions one practice trial was given (17 & 18) using a disjunctive question including some example answers so participants could become acquainted with the task.
(17) Amy en jij gaan vanavond uit eten. Ze wil graag een reservering maken. Ze vraagt je: ‘Wil je liever naar het pannenkoekenschip of naar de Chinees?’. (Je
wilt niet naar het pannenkoekenschip)
Wat antwoord je? bijvoorbeeld:
▪ 'Niet het pannenkoekenschip' ▪ 'Liever niet het pannenkoekenschip' ▪ 'Geen pannenkoeken!'
▪ 'Laten we naar de Chinees gaan' ▪ 'Liever de Chinees'
▪ 'De Chinees!’
Of iets anders. Antwoord zoals je in het écht ook zou antwoorden. English translation of (17):
(18) Amy and you are going out for dinner tonight. She wants to reserve a spot. She asks you: ‘Would you rather go and have pancakes or Chinese food?’. (you don’t want
to have pancakes)
For example:
▪ ‘No pancakes!’
▪ ‘No pancakes for me, please’ ▪ ‘Definitely no pancakes’ ▪ ‘Let’s go for Chinese food’ ▪ ‘I’d prefer Chinese food’ ▪ ‘Chinese!’
Or something else. Respond how you would in real life!
No feedback was given to the participants about their response. The example answers were kept as natural as possible to motivate and show the participants that they should really respond the way they would respond in real life, and that they did not have to give a full sentence answer. On average it took participants 22 minutes (range: 7 minutes to 1 hour and 39 minutes) to complete the task. Because it was an online task participants had to complete it within 24 hours. They could start the questionnaire, take a break and do something else before finishing it. This explains the wide range of time it took participants to complete the questionnaire. At the end of the test participants received a code they could send to the experimenter to be in the running to win three chocolate bars. If participants wanted more information they could proceed to the last page of the questionnaire for a debriefing (appendix H).
2.1.4 Data and Analysis
The given responses were analysed and categorized. The answers were divided into 13 different categories (19). Within these categories a distinction is made between bare polarity particles, and particles with an expanded reply. The expanded replies are categorized as either positive, negative, or neutral. Expanded replies are categorized as positive if they include a positive polarity particle such as wel, zeker or inderdaad. Expanded replies are categorized as negative if they includes a negative polarity item such as niet, geen or nooit. Finally, expanded replies are categorized as neutral if the reply does not contain any polarity item. The 13 different categories are:
(19) 1. nee 5. ja 9. jawel
2. nee + neutral 6. ja + neutral 10. jawel + neutral
3. nee + positive 7. ja + positive 11. jawel + positive
4. nee + negative 8. ja + negative 12. jawel + negative
13. other
The ‘other’ category included answers that did not contain either nee, ja or jawel, and ‘false’ answers, For which the answer was false, given the meaning to be expressed in the answer between brackets. For instance one participant used nee to answer the item: ‘a relative of your parents is visiting. He brought his dog. He asks you: ‘do you have any pets?’. (you have pets)’. Nee cannot be interpreted to mean that the participant has any pets, and is therefore marked as a ‘false’ response.
2.2 Results
In the following sections the results of the production experiment will be discussed. First, affirmative responses to positive polar questions will be discussed, then negative responses to positive polar questions, followed by affirmative responses to negative polar questions and finally, negative responses to negative polar questions.
2.2.1 Responding Affirmatively to a Positive Polar Question
In this section, the results will be presented of the control items to which participants had to respond affirmatively (see Appendix I for Table 10 including the absolute number of responses and percentages). The results indicate that nee is not used (0%) to affirmatively respond to positive polar questions (see Figure 5). To respond affirmatively to positive polar questions, participants
predominantly used ja (84.7%). Jawel is also used, once as a bare particle (0.6%) and twice with a neutral expanded reply (1,3%) by three different participants.
Ja as a bare particle was used in 32.1% of the cases. In 48.1% of the cases participants used a
neutral expanded reply (e.g., ‘Ja, die verkopen we.’, ‘Ja hoor’, ‘Ja, heb ik’). In 4.5% of all cases a positive expanded reply was used (e.g., ‘Ja, volgens mij wel.’, ‘Ja, gelukkig wel.’, ‘Ja, dit gaat zeker
