Previous talks


Modified numerals, ignorance inferences, and QUD effects

Alexandre Cremers

ILLC, University of Amsterdam

Modified numerals are known to give rise to an array of inferences, which have received considerable attention in the past few years: ignorance inferences in unembedded contexts, variation effects under universal quantifiers, upper-bound readings. An important debate focuses on the contrast between superlative (‘at least’) and comparative (‘more than’) modifiers. Experimental results seem contradicting as some studies found clear differences between comparatives and superlatives, while others found little to no evidence for a contrast.

In this study, we were primarily interested in how ignorance inferences are affected by QUD manipulations, and we tried to understand the conflicting results. We found that both ‘at least’ and ‘more than’ can give rise to ignorance inferences which are sensitive to the QUD. Interestingly, in truth-value and acceptability judgment tasks ‘more than’ was not found to give rise to any ignorance effect (in contrast with ‘at least’), while tasks which directly assess the inference showed similar ignorance inferences for both types of modifiers.

We also observed interesting effects with bare numerals, which overall support theories which derive their exact reading from an at-least denotation via pragmatic strengthening.



The processing effects of stance markers and connectives in on-line reading

Yipu Wei

Utrecht University

Background & Research questions

The sentence “The neighbours are not at home, because their lights are out”, contains a claim-argument relation. The connective because marks the fact that there is a causal relation. In the Dutch translation of the sentence the connective want ‘because’ not only expresses the causality of the relation, but it also specifies the fact that the first segment is a claim, and hence that there is a subjective relation between the segments (Spooren et al., 2010). On-line processing studies (Canestrelli et al., 2013) have shown that the subjective connective want leads to an immediate processing delay. This processing delay is cancelled out when a perspective/stance marker such as according to Peter is added to the first segment. Since the stance markers already marks the first clause as a subjective claim, the subjectivity information encoded by the connective want is not new to readers any more, and hence no delay follows.

In forward causal relations, such as “The neighbours’ lights are out, so they are not at home”, the connective so marks the second clause as subjective. In this case, a stance marker at the beginning of the first clause does alert the reader that reasoning is involved, but not specifically that the second segment is a claim. In the present paper we investigate what effect such a stance marker has on the processing of the subjectivity of the second segment. In addition, we investigate whether an attitudinal stance marker (e.g., fortunately; Conrad & Biber, 2000) has affects the processing of subjectivity in the same way as an epistemic stance marker (e.g., according to Peter).


We conducted an eye-tracking reading experiment in Chinese. The materials contained argument-claim relations with either the subjective connective kejian or the connective suoyi, which, like the English connective so is underspecified for subjectivity. We also varied the type of stance marking in the first segment (no stance marking, epistemic stance marker and attitudinal stance marker). A modal verb may/must was added in the second clause of the relation. The modal verb provided an unambiguous cue that the second segment contained a claim. The experiment was conducted with an EyeLink-1000 eye tracker.

Results & Conclusion

The subjective connective kejian led to a processing delay at the connective region compared to the underspecified connective suoyi, irrespective of the presence of stance markers. However, both epistemic stance markers and attitudinal stance markers facilitated the processing at the modal verb. Thus, the stance markers did facilitate the processing of subjectivity, but the subjectivity of the connectives still led to increased processing times initially.

Selected references

Canestrelli, A., Mak, W.M., & Sanders, T.J.M. (2013). Causal connectives in discourse processing: How differences in subjectivity are reflected in eye-movements. Language and Cognitive Processes, 28 (9): 1394-1413.

Conrad, S., & Biber, D. (2000). Adverbial marking of stance in speech and writing. In S. Hunston & Thompson,Geoffrey (Eds.), Evaluation in text: Authorial stance and the construction of discourse (pp. 57–73). Oxford and New York: Cambridge University Press.

Spooren, W.P.M.S., Sanders, T.J.M., Huiskes, M., & Degand, L. (2010). Subjectivity and causality: A corpus study of spoken language. In S. Rice & J. Newman (eds.), Empirical and experimental methods in cognitive/functional research (pp. 241-255). Stanford: CSLI Publications.



The psycholinguistics and neurolinguistics of regional accent processing

Cesko Voeten

Leiden University

Large-scale cross-sectional investigations have shown that Standard Dutch (spoken in the Randstad area) and varieties of Flemish differ in various phonetic respects. Salient differences are found in the vocalic domain, for example with respect to the realizations of the tense mid vowels and diphthongs (e.g. [1,2,3]), and in the consonantal domain, e.g. in the allophone inventory of the rhotic ([4]). The present research focuses on three sets of differences, which all happen to be due to sound changes that have occurred in the Netherlands but not in Flanders. These are the diphthongal (Standard Dutch) versus monophthongal (Flemish) realizations of /e:,ø:,o:/, the concomitant stronger diphthongization of (ɛi,œy,ɑu) in Standard Dutch, and the double allophones [ʀ~ɹ] (Standard Dutch) versus single allophone [r] (Flemish) for the rhotic phoneme.

At the level of the individual the effect of these accentual differences on inter-accentual processing has not received a lot of attention. Studies on accent processing in general have found subtle processing effects in perception, such as attenuated N400 ERPs ([5]) and slightly longer reaction times to auditory words (up to 30 ms; [6]), which persist even when the listener receives more exposure ([7]). Production studies have shown that adjustments of a person’s speech can occur ([8]), but not through simple imitation ([9]).

The present study takes a closer look at the processing of Standard Dutch versus Flemish realizations of the sounds and sound contrasts discussed in the first paragraph, by Flemish first-year students in the Netherlands and Randstad Dutch controls. Three sources of evidence will be discussed: speech production data, behavioral perception data, and electrophysiological evidence. Taken together, these three sources aim to provide converging evidence on both phonetic adaptation to accent differences in general, and, more specifically, on the malleability of allophonic representations. These types of data are couched within a larger longitudinal experiment (the current presentation focuses on the first session of three), aiming to shed light on the adaptation that will take place in the perception and production of these Flemish students to the Standard Dutch accent.


[1] Van de Velde, H. (1996). Variatie en verandering in het gesproken Standaard-Nederlands (1935-1993). PhD dissertation, Radboud University Nijmegen.

[2] Adank, P., van Hout, R., & Smits, R. (2004). An acoustic description of the vowels of Northern and Southern Standard Dutch. The Journal of the Acoustical Society of America, 116(3), 1729–1738.

[3] Adank, P., van Hout, R., & Van de Velde, H. (2007). An acoustic description of the vowels of Northern and Southern Standard Dutch II: regional varieties). The Journal of the Acoustical Society of America 121(2), pp. 1130–1141.

[4] Sebregts, K. (2015). The sociophonetics and phonology of Dutch r. Utrecht: LOT.

[5] Goslin, J., Duffy, H., & Floccia, C. (2012). An ERP investigation of regional and foreign accent processing. Brain and language, 122(2), 92-102.

[6] Floccia, C., Goslin, J., Girard, F., & Konopczynski, G. (2006). Does a regional accent perturb speech processing? Journal of Experimental Psychology: Human Perception and Performance, 32(5), 1276.

[7] Floccia, C., Butler, J., Goslin, J., & Ellis, L. (2009). Regional and foreign accent processing in English: Can listeners adapt? Journal of Psycholinguistic Research, 38(4), 379-412.

[8] Pardo, J. S., Gibbons, R., Suppes, A., & Krauss, R. M. (2012). Phonetic convergence in college roommates. Journal of Phonetics, 40(1), 190-197.

[9] Pardo, J. S. (2012). Reflections on phonetic convergence: speech perception does not mirror speech production. Language and Linguistics Compass, 6(12), 753-767.



Information content of functional networks as a correlate of conscious awareness

Ramon Guevara Erra 

Université Paris Descartes

It is said that complexity lies between order and disorder. In the case of brain activity and physiology in general, complexity issues are being considered with increased emphasis. We sought to identify features of brain organization that are optimal for sensory processing, and that may guide the emergence of cognition and consciousness, by analyzing neurophysiological recordings in conscious and unconscious states. We find a surprisingly simple result: Normal wakeful states are characterized by the greatest number of possible configurations of interactions between brain networks, representing highest entropy values. Therefore, the information content is larger in the network associated to conscious states, suggesting that consciousness could be the result of an optimization of information processing.



How Cantonese-speaking adults use prosody to interpret focus in L2 English: Evidence from eye movement in the visual world paradigm

Emily Haoyan Ge

Chinese University of Hong Kong

This study investigated whether scalar implicatures can be derived for sentences with modified numerals (e.g. at least n and more than n) embedded under disjunctive conjunction (e.g. both…and…). Various studies suggested that theoretically, the implicatures should be available, but only a few studies tested them experimentally. An experiment was designed in this study to detect this type of scalar implicature. Covered box paradigm was chosen as the methodology. The results of the experiment show a very low implicature derivation rate (1.28%). Instead of claiming that scalar implicatures are unavailable for modified numerals, I first discussed the potential problems in experimental design and then provided the possible explanations for the infrequent implicature derivation of modified numerals.

This eye-tracking study uses the visual world paradigm to investigate how Cantonese-English bilingual adults integrate prosody and other domains of linguistic knowledge in the interpretation of focus in English. Focus in English is typically realized by assigning an accent to the focal element. In sentences with the focus particle only, different accent placement triggers different sets of alternatives and affects the truth-value of the sentence (Jackendoff, 1972; Rooth, 1992), as in (1). By contrast, the use of prosody to realize focus in Cantonese is highly constrained, as the pitch range is used for lexical contrasts (Chao, 1947)

(1)  The dinosaur is only carrying the bucket.

  1. The dinosaur is only carrying the BUCKETnot carrying the suitcase.
  2. The dinosaur is only CARRYING the bucket, not throwing the bucket.

Forty Cantonese-English bilingual adults and forty native speakers of English participated in this study. They heard Englishonly-sentences with the accent falls on either the object or the verb while looking at four pictures. By measuring the time course of eye movements, the study aims to detect the earliest point that participants’ fixation patterns give evidence that they consider the alternativeA post-test acoustic task was conducted to examine whether participants are able to detect the placement of accent in speech perception.

Native controls performed anticipatory eye-movements to the alternatives before the offset of not based on the presence of prosody, whereas Cantonese speakers didn’t fixate more on the alternatives until hearing the verb following not. In the acoustic task, both groups demonstrated around 90% accuracy rate and the reaction time didn’t reveal any significance between groups. The results suggest that though bilingual adults are able to detect accents, they have difficulty in integrating prosodic cues and other levels of linguistic knowledge.



Chao, Y. R. (1947) Cantonese Primer. Cambridge, Massachusetts: Harvard University Press.

Jackendoff, R. S. (1972) Semantic interpretation in generative grammar. Cambridge, MA: MIT Press.

Rooth, M. (1992) A theory of focus interpretation. Natural Language Semantics 1. 75–116.



Interpreting Lower-bound Modified Numerals under Distributive Conjunction  

Ying Liu 

Utrecht University

This study investigated whether scalar implicatures can be derived for sentences with modified numerals (e.g. at least n and more than n) embedded under disjunctive conjunction (e.g. both…and…). Various studies suggested that theoretically, the implicatures should be available, but only a few studies tested them experimentally. An experiment was designed in this study to detect this type of scalar implicature. Covered box paradigm was chosen as the methodology. The results of the experiment show a very low implicature derivation rate (1.28%). Instead of claiming that scalar implicatures are unavailable for modified numerals, I first discussed the potential problems in experimental design and then provided the possible explanations for the infrequent implicature derivation of modified numerals.



Specificity and entropy reduction in situated referential processing 

Elli Tourtouri 

Saarland University

In situated communication, reference to an entity in the shared visual context can be established using either an expression that conveys precise (minimally specified) or redundant (over-specified) information. There is, however, a long-lasting debate in psycholinguistics concerning whether the latter hinders referential processing (e.g., Arts et al., 2011; Engelhardt et al, 2011; Tourtouri et al. 2015). I will present eye tracking data showing that over-specifications in fact aid listeners in their effort to identify the visual target. Furthermore, I will present evidence that, above and beyond any effects of specificity, referring expressions that maximally reduce the referential search space early on also benefit processing.



Existential verbs, article drop and definiteness

Bert Le Bruyn

Utrecht University

The peculiar interaction between verbs like have, give, wear, etc. and their objects has recently been noted in work on definiteness and incorporation. Coppock & Bevaar (2015) show how combining these existential verbs with the only N gives rise to non-presuppositional readings of the definite article ((1) vs. (2)). Le Bruyn, de Swart & Zwarts (2016) further show how these verbs are more likely than others to allow for indefinite article drop in article languages ((3) vs. (4)).
(1) Mary didn’t hear the only brilliant talk. (only one brilliant talk)
(2) Mary didn’t give the only brilliant talk. (more than one brilliant talk)
(3) María ví *(un) sombrero. (‘Mary saw (a) hat’) SPANISH
(4) María llevaba (un) sombrero. (‘Mary wore (a) hat’) SPANISH
CB and LSZ’s work focuses on Romance and Germanic. We claim that existential verbs are special at a broader cognitive level and argue for this by demonstrating the predictive power of the existential vs. regular verb distinction in the L2 English article acquisition of Chinese native speakers, uniquely integrating article drop and overproduction. Our argumentation will be based on data from the Cambridge Learner Corpus.



The relation between linguistic skills and problem behavior in preschoolers

Loes Janssen, Brigitta Keij & Martjin de Ouden

Utrecht University

Previous research has demonstrated a high co-occurrence of linguistic problems and behavior problems in children. Whether linguistic problems are a determinant of behavior problems in typically developing children remains unclear. This study examined whether linguistic skills could predict problem behavior in preschoolers. By evoking ‘communication breakdown’ at different linguistic levels during an innovative, interactive tablet game, we studied the influence of linguistic skills on problem behavior using an new coding system: System for Coding Child Behavior in Interactive Tasks. In total, 50 monolingual children (4-6 years) have participated in the study. The PPVT scores from a subsample (= 50) validated the game as a measure of the children’s linguistic abilities, (50) = .503, < .001. Teacher reported child behavior on a behavioral questionnaire (SDQ) validated the intensity of the coded behavior, r (37) = .339, p = .040. The total score of the game shows a negative correlation with the total intensity of child problem behavior, when excluding nervousness, r (50) = -.371, p = .008. In conclusion, the results indicate that children who encountered more communication problems in the game demonstrated more intense problem behavior than children who experienced less communication problems in the game. The early identification of linguistic problems is of great importance for timely intervention, because specific underlying issues in communication could be targeted before behavior problems develop.



Phoneme categorization, behavioral and EEG measurements

Anne-France Pinget & Ao Chen  (Utrecht University)

Part 1: Anne-France Pinget

Regional accent effects on speech perception

Perception of phonetic detail is dynamic, malleable and listener-specific. Many social factors have recently been shown to play a role in speech perception (f.i. age, ethnicity, regional accent, gender, etc.). In this talk, I discuss the effects of regional accent on speech perception more specifically. Both listener’s and speaker’s oriented aspects are reviewed.

In the first part, I summarize the conclusions reached in my doctoral thesis based on behavioral data, where regional differences were found in the perception of a consonantal contrast. These differences in speech perception could also be related to differences in speech production, which provides evidence for an abstract link between the production and perception systems.

In the second part, I discuss current and future research plans in which regional effects on speech perception are investigated by means of EEG data.


Part 2: Ao Chen

Interaction between phoneme categorization and word learning

Both phoneme categorization and word learning have been widely studied in infant research, yet barely any study has investigated the interaction between the two. The VENI project will examine brain responses (event-related potentials, ERPs) of 20-month-old typically developing children and children at familial risk of dyslexia on both phonemic categorization (PC) and word learning (WL). Alongside the online word learning experiment, infants’ overall word knowledge will be examined using the Dutch version MacArthur-Bates Communicative Inventory (N-CDI). If PC contributes to learning new minimally different words, then the initial PC should correlate with word learning success. If WL enhances PC in return, then PC should improve after learning the words. Children at familial risk will be compared with typically developing children on PC before and after WL, WL, and overall vocabulary.

In the talk, I will present 1) the design of the whole project, 2) preliminary results of PC among adult participants, which will serve as the baseline for the infant experiments, 3) design of infant experiments.



Exploring the use of automatic subjectivity analyses for Spanish causal connectives in various text types

Andrea Santana Covarrubias (joint work with Dorien Nieuwenhuijsen, Wilbert Spooren & Ted Sanders)

1 Utrecht Institute of Linguistics (UiL OTS), 2 Radboud Universiteit

Causality and subjectivity are relevant cognitive principles in the categorization of coherence relations and connectives. Studies in various languages have shown how both principles are encoded by different connectives. However, the Spanish language has been understudied from this perspective. Also, most of the existing research on connectives has used manual analyses, mainly because of the complexity involved in the analysis of discourse relations. In this paper, we explore the use of automatic analyses of subjectivity in causal connectives. The goal is to determine the degree to which Spanish causal connectives encode subjectivity across different text types, by carrying out automatic analyses. Our assumption is that if a connective signals subjectivity, it will occur in a subjective environment, that is, a context containing relatively many subjective words. Therefore, we constructed a corpus and identified causal connectives in journalistic discourse (news and editorials) and academic discourse (essays, research articles and textbooks of Education and Psychology). In order to automatically identify the frequency of subjective words in the texts and the segments linked by the most frequent causal connectives we used a Spanish lexicon of subjectivity[1]. In this study, we show that there is a statistically significant relationship between the text type and subjectivity. As far as methodology is concerned, we also discuss the advantages and problems of the use of automatic analyses. We argue that the combination of automatic and manual analyses can result in a promising methodology for the study of discourse coherence.


Molina-González, M.D., Martínez-Cámara, E., Martín-Valdivia, M.T., Perea-Ortega, J.M. 2013. Semantic orientation for polarity classification in Spanish reviews. Expert Systems with Applications 40-18: 7250–7257.



Quantity implicatures and autism spectrum disorders

Bob van Tiel (ULB)

(joint work with Mikhail Kissine)

We investigated how frequently participants with and without autism spectrum disorders derived four types of pragmatic inferences, each of which involves reasoning about what the speaker could have said. In line with previous research, we show that the probability of deriving simple pragmatic inferences—for which one only has to reason about the meanings of the words the speaker could have said—is independent of one’s position on the autism spectrum. However, if the derivation of the pragmatic inference involves reasoning about what the speaker would have implied had she made a different utterance, the probability that the pragmatic inference is derived decreases significantly with one’s autism spectrum quotient. We discuss the consequences of our findings for theories of pragmatics in autism and for linguistic theorising in general.



Individual Differences in Study Abroad: Oral Fluency, the Study Abroad Experience and Vocabulary Development 

Sapna Sehgal  (University of Barcelona) 

While it is generally accepted that the stay abroad (SA) context positively affects language learning, especially in the area of oral fluency, we do not yet have a full picture of the role individual differences (IDs) play in speaking fluency development. Recent studies have emphasized the importance of cognitive abilities such as inhibitory control on speaking fluency (Segalowitz 2016, 2010), and the need to precisely measure disfluencies (silent and filled pauses, repetitions, repairs, etc.) in second language (L2) speech (e.g. de Jong, 2016, Prefontaine & Kormos, 2016). The current study investigates differences in inhibitory control and the study abroad experience on L2 fluency.
52 American participants studying in Barcelona for 3 months completed inhibitory control, speaking, vocabulary and questionnaire tasks at pre and post-test. A Simon task and a picture decision task (Colomé, 2001) were used to measure inhibitory control. Fluency data was taken from speaking tasks; additional language switching tasks from L1-L2 and L2-L1 provided data on switching fluency. The discussion will involve a preliminary analysis of fluency data conducted thus far. It will focus on vocabulary development and questionnaire data, including the frequency of L2 use. While group results show no significant difference in inhibitory control, individual analysis show a few participants who are potentially better inhibitors, and some participants showed a significant difference in oral fluency from pre to post test. Individual gains in vocabulary development were also shown for many participants, though most participants reported using the L2 less than 50% of the time during their stay abroad.



To each their own: Investigating the Dutch masculine genericzijn ‘his’ across stereotype contexts

Theresa Redl  (Utrecht University)

Discourse comprehension can be seen as the successful construction of a mental model of the situation described in a discourse. This mental model or situation model comprises certain information about the protagonists, including their gender (Zwaan & Radvansky, 1998). Language users have been shown to infer the gender of protagonists when it is not explicitly mentioned (e.g., Carreiras et al., 1996). Such gender inferences can be based on at least two types of gender cues: stereotypes and masculine generics. For example, in (i) language users are likely to infer the protagonist’s gender based on the stereotype that secretaries are female:

  1. Miss Smith asked her secretary to take a letter. His pencil kept breaking.

Furthermore, gender inferences can be based on the grammatical gender of masculine generics. Masculine generics are grammatically masculine terms that are used to refer to people of unknown, unspecified or mixed gender (Braun et al., 2005). Thus, this covers pronouns like his as in the proverb ‘To each his own’, but also role names in languages like German and French where the masculine noun is used as the default (e.g., German die Studenten ‘the students, masc.’). Recent research combining role names as masculine generics and stereotypes has suggested that masculine generics induce a male bias regardless of the context in which they occur (e.g., Gygax et al., 2008).

In an eye-tracking experiment with text stimuli, I extended this recent line of research to Dutch. I addressed the question whether the Dutch masculine generic zijn ‘his’ is interpreted generically, or whether it induces a male bias. In order to test the influence of context, zijn ‘his’ was presented in stereotypically female, male and neutral contexts. Consider the following text featuring a male stereotype for illustration:

  1. Iedereen was zijn autobanden aan het verwisselen. Zo was ook Lieke/Peter weer de zomerbanden onder de auto aan het zetten.
    ‘Everyone was changing his tires. Lieke/Peter, too, was installing summer tires on the car.’

The results based on the data of 62 participants did not support the hypothesis that zijn‘his’ induces a male bias. Thus, zijn ‘his’ was interpreted as intended, namely generically. Gender inferences were instead based on stereotypes. Participants faced processing difficulties when the protagonist’s gender did not match the stereotype, but intriguingly this effect was only found for male protagonists. Thus, reading about a man indulging in a stereotypically female activity increased reading times, while this effect was not significant for female protagonists indulging in stereotypically male activities. The results will be discussed in the light of research from social psychology on gender role violations.


Braun, F., Sczesny, S., & Stahlberg, D. (2005). Cognitive effects of masculine generics in German: An overview of empirical findings. Communications30(1), 1–21.

Carreiras, M., Garnham, A., Oakhill, J., & Cain, K. (1996). The use of stereotype gender information in constructing a mental model: Evidence from English and Spanish. Quarterly Journal of Experimental Psychology49A(3), 639–663.

Gygax, P., Gabriel, U., Sarrasin, O., Oakhill, J., & Garnham, A. (2008). Generically intended, but specifically interpreted: When beauticians, musicians, and mechanics are all men.Language and Cognitive Processes23(3), 464–485.

Zwaan, R., & Radvansky, G. (1998). Situation models in language comprehension and memory. Psychological Bulletin123(2), 162–185.



Contact-induced phonological emergence in Dutch

Alexander Martin (École Normale Supérieure)

(joint work with Marieke van Heugten, René Kager, Sharon Peperkamp)

Dutch is undergoing an interesting change in its phonological system. Through contact with English and the borrowing of a great number of words, the sound /ɡ/, formerly absent from the language, is becoming contrastive (e.g., /koːl/, cabbage ~ /ɡoːl/, goal). In the present study, we examine both the production and perception of the emerging /k/~/ɡ/ contrast, specifically targeting regional differences between the northern and southern parts of the Dutch-speaking area of Europe. We found higher rates of /ɡ/ production in the North than in the South, and also that perception correlated with production in the South, but not the North. Discussion will focus on the social factors that might influence this finding.



The prosody of wh-indeterminates in Mandarin and its implications on prosodic phrasing

Yang Yang (Leiden University)

In Mandarin, wh-words can have an interrogative and a non-interrogative interpretation. A sentence containing diǎnr (a little) and shénme ‘what’ can be interpreted either as a question or as a declarative sentence, with shénme interpreted either as ‘what’ or as an indefinite meaning ‘something’. In order to investigate the prosodic properties of these two sentence types and their prosodic phrasings, we conducted a production study (34 participants) on above structure by designing two conditions, wh-questions and wh-declaratives. The two conditions are identical except the punctuation at the end. Our study clearly showed that there are different prosodic cues in wh-questions and wh-declaratives. Importantly, it is duration instead of F0 that provides an early cue from the onset of the utterance. This study also implies that declaratives that contain shénme differ from questions in terms of prosodic phrasing.



The semantics of L2 definites: the meaningful bounds of specificity

Bert Le Bruyn & Xiaoli Dong (Utrecht University)

The SLA and the semantics literature on definites have rich but largely disjoint traditions. Recent semantic literature focuses on uniqueness and familiarity as key ingredients of definites across languages (a.o. Lyons 1999; Schwarz 2009; Arkoh & Matthewson 2013) and tries to figure out how the two interact in languages like English (a.o. Farkas 2002; Roberts 2003). Recent SLA literature and in particular the one looking at the acquisition of definites by learners with an articleless L1 has focused on specificity (a.o. Ionin 2003; Ionin et al. 2004; Tryzna 2009; Yang & Ionin 2009; Snape, Leung & Ting 2006; García-Mayo & Hawkins 2009; though see Trenkic et al. 2014). We argue that Mandarin learners of English correctly use and interpret definites that are familiar and unique but that they have problems as soon as we move to more complex cases in which the definite is unique but not familiar. We further show that these complex cases are exactly those in which current semantic approaches fail to make the right predictions and explore the relevance of the notion of specificity in light of these new findings.



Surface vs. Underlying Listening Strategies for Cross-Language Listeners in the Perception of Sandhied Tones in the Nanjing Dialect

Xin Li & René Kager  (Utrecht University)

This study is devoted to explore surface/underling listening strategies adopted by native and non-native, tone and non-tone language groups in their perception of sandhied tones, and the possible influence from phonetic naturalness of the sandhi rule on the perception. The mapping between the sandhied tones and their surface/underlying tones are compared in two Nanjing tone sandhi rules differing in phonetic naturalness, across Dutch group, Beijing group and the native Nanjing group through a Concept Formation paradigm. Results reveal distinct listening patterns in the three language groups. Dutch listeners experience difficulty in creating phonological representation for tones even in surface listening; Beijing listeners apply their native grammar of tones in interpreting Nanjing sandhied tones in surface listening but have no access to their underlying patterns; Nanjing listeners listen to their native sandhied tones at both the underlying level and surface level and they seem to always mix the two levels when processing tones. Phonetic naturalness of the sandhi rule does not seem to play a role in the current experiment.



Embodying Emotion in Narrative Discourse: a facial EMG study on simulation vs. moral evaluation in affective language processing

Björn ’t Hart(joint work with Marijn Struiksma1, Anton van Boxtel2, Jos van Berkum1)

1: Utrecht University, 2: Tilburg University

Facial electromyography (fEMG) research has shown that activation and deactivation of the corrugator supercilii (‘frowning muscle’) reliably indexes negative and positive affect respectively. Furthermore, work on embodiment in language processing suggests such muscle activity is involved, perhaps even causally, in conceptual simulation during language comprehension.  So far, this work focused mainly on simple words and sentences (‘happy’, ‘John was happy’), which provide affectively valenced lexical concepts and situation models to be simulated, but not much else. In natural language use, however, people usually also affectively evaluate things. Importantly, corrugator activity involved in evaluation might be at odds with corrugator activity implicated in simulating the described concepts or events, e.g. when an immoral character enjoys something positive. To investigate the extent to which, in more complex situations, corrugator activity reflects simulation (lexical/situation model valence) and/or moral evaluation (reader appraisal), we asked participants to read 64 narratives that orthogonally manipulated valence at both levels. Reading about characters behaving in a morally laudable or objectionable fashion in the first half of the narrative immediately led to decreased or increased corrugator activity respectively. Critically, subsequent events befalling these good or bad characters elicited corrugator activity demonstrating an interaction between simulation and moral evaluation: positive and negative events befalling negatively evaluated characters shifted corrugator activity towards the opposite valence. This suggests that neither simulation nor moral evaluation alone determine embodied fEMG activity, and that both involve facial muscle activity during online language comprehension. This highlights the importance of considering evaluative stance in embodied language processing.



The role of metacognitive knowledge in processing coherence markers

Rianne Vlaar  (Utrecht University)

Coherence markers such as connectives have often been found to positively influence both the reading process and reading comprehension (e.g. Millis & Just, 1994; Sanders & Noordman, 2000). However, this positive effect does not seem to hold for all individuals. Knowledge of connectives is a predictor of text comprehension above and beyond reading fluency, general vocabulary knowledge and metacognitive knowledge (Welie, Schoonen, & Kuiken, 2016). Moreover, knowledge of connectives interacts with metacognitive knowledge – the ability to use strategies to regulate the reading process. Metacognitive knowledge appears to mediate the extent to which readers can make use of their knowledge of connectives. The present study will investigate how this mediating effect of metacognitive knowledge on knowledge of connectives influences the reading process and reading comprehension. In this talk, the background, design and hypotheses of the study will be discussed. As the study will be carried out later this spring, feedback on the design is very welcome.


Millis, K. K., & Just, M. A. (1994). The influence of connectives on sentence comprehension. Journal of Memory and Language,33(1), 128-147.

Sanders, T. J., & Noordman, L. G. (2000). The role of coherence relations and their linguistic markers in text processing.Discourse processes29(1), 37-60.

Welie, C., Schoonen, R., & Kuiken, F. (2016). Expository text comprehension: for which readers does knowledge of connectives contribute the most? Manuscript submitted for publication.



The processing of subjectivity in coherence relations: evidence from the visual world paradigm

Yipu Wei (Utrecht University)

The processing of discourse includes the interpretation of coherence relations between utterances. These relations may be based on objective facts (objective relation) or originate from some intentional mind (subjective relation). In establishing a subjective relation, the recognition of who is doing the reasoning, i.e. theSubject of Consciousness (SoC; Pander Maat & Sanders, 2001; Sanders, Sanders & Sweetser, 2012) is crucial. Connectives, as linguistic devices of coherence relations, encode different degrees of subjectivity. Some connectives are prototypical for subjective relations, while some are specific for objective ones. Some general connectives can be used in both ways. In this study, we aim to investigate whether the degree of subjectivity of a connective directs people’s attention in identifying the SoC.
We conducted a visual world paradigm eye-tracking experiment with the EyeLink-1000. In the experimental setting, participants listened to sentences while they were presented with two scenes. One scene depicted the event being described by the first clause in the auditory input, the other depicted the SoC. The latter picture involved someone speaking, with a speech balloon in which the event was visible. Participants heard sentences connected by either the subjective connective or the objective one. We tested the effect of subjective and objective connectives in both Dutch and Chinese.
The Dutch result suggested a significant increase in the proportion of looks on the with-SoC scene after the subjective connective dus ‘so’ compared to the objective connective daardoor ‘as a result’. In Mandarin Chinese, the subjective connectives kejian ‘so’ behaved similarly to the generic connective suoyi ‘so’ (can be used either subjectively or objectively): both of these two connectives led to an increase of looks on the with-SoC scene. However, the objective connective yin’er ‘as a result’ guided the attention away from the with-SoC scene. It can be inferred from this study that the interpretation of the subjective relation does involve a process of identifying the SoC that is responsible for the subjectivity. Moreover, the subjectivity of the connectives function as linguistic cue which directed language users’ attention to the responsible subject.


Canestrelli, A. R., Mak, W. M., & Sanders, T. J. M. (2013). Causal connectives in discourse processing: How differences in subjectivity are reflected in eye-movements. Language and Cognitive Processes28(9), 1394–1413.
Li, F. (2014). Subjectivity in Mandarin Chinese: The meaning and use of causal connectives in written discourse. Utrecht: LOT.
Pander Maat, H. & Sanders, T(2001). Subjectivity in causal connectives: An empirical study of language in use. Cognitive Linguistics, 12, 247273.
Sanders, J., Sanders, T., & Sweetser, E(2012). Responsible subjects and discourse causality. How mental spaces and perspective help identifying subjectivity in Dutch backward causal connectives. Journal of Pragmatics, 44(3)169-190.



The Developmental Trajectories of Attention Distribution and Segment-Tone Integration in Dutch Learners of Mandarin Tones

Ting Zou (Leiden University)

This study investigates how beginner and advanced Dutch learners of Mandarin process tonal information. An ABX matching to sample task is adopted to investigate phonological discrimination of Mandarin tones and segment-tone integration in Dutch learners of Mandarin, with both native Mandarin and Dutch speakers (without tonal learning experience) as control. Results show a developmental path in tone learning. The beginner learners cannot process tonal contrast adequately at phonological level, and they process segmental and tonal information separately, like native Dutch listeners without Mandarin experience. The advanced learners showed a good phonological discrimination of tonal contrasts. They show a more native-like pattern in distributing their attention between segmental and tonal information, and they process the two dimensions in an integrated manner, similar to native Mandarin listeners. This suggests that the acquisition of new tonal categories in L2 involves a redistribution of attention along acoustic dimensions and the development of segment-tone integration.

Keywords: lexical tone learning, attention redistribution, segment-tone integration, Dutch learners of Mandarin.



Can Scalar Implicatures be drawn from Modified Numerals?

Erlinde Meertens (Utrecht University)

This study discusses the availability of Scalar Implicatures (SIs) for modified numerals (more than / at least n). It is known that unembedded modified numerals never trigger SIs, but it is unclear if an SI is available when the modified numeral is embedded under a universal quantifier (2). 

1. Barbara read more than 9 books.

SI interpretation: Barbara read exactly 10 books.                            X

2. Every girl read more than 9 books.

SI interpretation: Some student read exactly 10 books.                  ✔

Mayr (2015) predicts that an SI is available for both comparative (more than) and superlative (at least) modifiers. Whilst others (Nouwen,2008;  Fox and Hackl, 2006) have arguments contradicting this account. This study investigates the availability of SIs for modified numerals in an truth-value judgment task with reaction time measurements following Bott and Noveck (2004). We found that in our experiment no SIs were drawn from modified numerals, contradicting the predictions of both Mayr (2015), Nouwen (2008) and Fox and Hackl (2004).  


Bott,  L.,  and  Noveck,  I.  (2004).  Some  Utterances  are  underinformative:  The  onset  and  time  course  of scalar inferences. Journal of Memory and Language 51:437-457.

Fox, D., & Hackl, M. (2006). The universal density of measurement. Linguistics and Philosophy, 29(5), 537-586.Hackl, M. (2000). Comparative quantifiers, Doctoral dissertation, MIT.

Mayr,  C.  (2015) Implicatures  of  modified  numerals.  In  I.  Caponigro  and  C.  Cecchetto  (ed.)  From Grammar  to  Meaning,  The  Spontaneous  Logicality  of  Language.  Cambridge  University  Press. 139-171.

Nouwen,  R.  (2008).  Upperbounded  no  more:  the  exhaustive  interpretation  of  non-strict  comparison. Natural Language Semantics, 16(4): 271-95.




Silvia Rădulescu (Utrecht Universtity)

In language acquisition, children manage impressively fast to infer generalized rules from a limited set of linguistic items, and apply those rules to novel strings. This research investigates what triggers and what limits the inductive leap from memorizing specific items to extracting abstract rules that apply productively beyond those items. Our new entropy model predicts that generalization is a cognitive mechanism that results from the interaction of input complexity (entropy) and brain’s limited processing and memory capacity (i.e. limited channel capacity).

            It was argued that children detect patterns in auditory input, like phonotactic information (Chambers, Onishi & Fisher, 2003), and word boundaries (Saffran, Aslin & Newport, 1996) by statistical learning. Statistical learning deals with computing probabilities that specific items co-occur, thus it cannot account for abstractions beyond those items. Gómez & Gerken (2000) drew a distinction between abstractions based on specific items (e.g. ba follows ba) and category-based abstractions (generalizing beyond specific items, to form abstract rules, e.g. Noun-Verb constructions). An algebraic system was proposed (Marcus, Vijayan, Rao & Vishton, 1999) to account for extracting rules that apply to categories, such as “the first item is the same as the third item” (li_na_li). This system addresses abstractions to novel items, but it does not explain how humans tune into such algebraic rules, and what the factors (if any) in the input are that facilitate or impede this process. Our entropy model addresses these questions and bridges the gap between previous findings, thus unifying them under one consistent account. According to our model, less complexity in the input facilitates memorization of specific items, which allows for abstractions based on those items, while a higher input complexity that overloads the channel capacity drives the tendency to make category-based generalizations (i.e. reduce the number of features that items can be coded for, and group them in abstract categories and acquire relations between these categories).

            We tested the predictions of this model in two artificial grammar experiments with adults. The effect of input complexity (entropy) on rule induction was tested in six conditions (different degrees of input complexity). As predicted, the results showed that the higher the input complexity, the higher the tendency to abstract away from specific items and to make category-based generalizations. Unlike previous proposals, this model also gives a quantitative measure for the likelihood of making generalizations in different ranges of input complexity. To further test our model and its domain generality, similar studies will be run with infants, and also using visual input.



Hitting playfully and kissing angrily: A self-paced reading study on force inferences

Suzanne Dekker & Anja Goldschmidt (UiL OTS)

Despite a lot of work done in the domain of verb-adverb modification (cf. e.g. Parsons, 1990; Eckardt, 1998; Piñón, 2007), not all meaning aspects arising in the process of modification have been accounted for. One class of such understudied aspects of modification are meanings that arise as the result of an inference. For instance, in (1), one would assume that Nancy hit Oliver with little force. This is confirmed by (2): The sentence sounds odd if we contrast playfully with lightly, suggesting that these adverbs’ meaning contributions are similar.

(1) Nancy hit Oliver playfully on the arm.
(2) Nancy hit Oliver playfully, but still rather ?lightly/ühard, on the arm.
(3) Nancy hit Oliver angrily, but still rather ?hard/ülightly, on the arm.

Yet the “force reduction” reading of (1) is not due to the lexical semantics of the adverb playfully, as this reading can be cancelled, cf. in combination with hard in (2). Similarly, for angrily in (3), a “force increase” inference arises (cf. incompatibility with hard), which can again be cancelled (compatibility with lightly). This is not an individual occurrence, but a systematic pattern, cross-cutting traditional distinctions between e.g. manner adverbs, mental-attitude adverbs, resultatives and subject depictives.

We have tested the strength of these inferences in a self-paced reading experiment. 72 adult Dutch native speakers read 30 test sentences and 30 control sentences in a Latin square design. Following the reading task, participants were asked to rate each sentence on a 4-point Likert scale (1 – clearly bad, 4 – clearly good).

The results show a significant reading time delay (p<0.05) on the adverb that cancels the inference in the stimuli condition and on the second and third word of the spill-over area (force increase condition), as well as on the first and second word of the spill-over area (force decrease condition).

Our experiment thus shows that this type of inference is not trivial, but something that theories about modification should explain.


Eckardt, Regine. 1998. Adverbs, Events, and Other Things. Tübingen: Max Niemeyer Verlag.
Parsons, Terence. 1990. Events in the Semantics of English. Cambridge, MA: MIT Press.
Piñón, Christopher. 2007. Manner adverbs and manners. Handout at the 7. Ereignissemantik-Konferenz, Schloss Hohentübingen.



Focus association revealed in reading times

Barbara Tomaszewicz (University of Cologne)

A focus associator is an expression whose contribution to the meaning of a sentence depends on the position of sentence focus. The focus structure introduces a presupposition about the context (Rooth 1992, 1996) and this presupposition can affect the setting of the domain restriction of quantificational expressions such as onlyevenalways (Rooth 1992, 1996, von Fintel 1994). Various semantic and syntactic diagnostics indicate that there are two types of focus associators: those that obligatorily require focus and those that freely/optionally associate with narrow focus (Beaver and Clark 2008).

How do the two types of focus-associating expressions affect expectations in online sentence processing? Obligatory associators can be predicted to create an expectation for the presence of focus in their scope during incremental semantic processing. But what happens in the case of optional associators? Our experiments indicate that both obligatory and optional associators create a processing bias for narrow focus indicating that optional focus association is not on par with contextual domain restriction of quantifiers (von Fintel 1994), i.e. the processing of a set of focus alternatives is lexically triggered and not merely the result of the fact that the restrictor variable tends to be resolved to focus alternatives that are contextually salient.



A new method for language comprehension reveals better performance on passive and principle B constructions

Shalom Zuckerman, Manuela Pinto, Elly Koutamanis and Yoin van Spijk , Utrecht University

See this link



Learnability of fusional and agglutinative morphology: Insights from artificial language learning

Alexis Dimitriadis (joint work with Natalie Boll-Avetisyan & Tom Fritzsche – University of Potsdam)

The opposition of fusional and agglutinative morphological systems, one of the oldest concepts in language typology and comparative linguistics, is often viewed in terms of learnability: In general, “transparent” agglutinative morphology (e.g., in Hungarian) should be easier to acquire than opaque fusional morphology (e.g., in German); cf. Pinker 1996, Zuidema 2003, Neeleman and Szendrői 2005, Fasanella 2014). However, until now such claims have not been supported by evidence from human learning behavior.

The artificial language learning paradigm (Reber, 1967) is ideal for testing these claims in highly controlled settings. In our study, subjects learn either a fusional or an agglutinative artificial language consisting of 360 morphologically complex nonwords marked for number and gender. In order to assess potential effects of language experience on learnability, we tested native speakers of German and Hungarian on both language types. Participants heard a nonword and simultaneously saw two pictures on screen. Their task was to decide which of the two pictures the word referred to.

Starting at chance, participants reached accuracy rates of over 90% in the course of the experiment. The agglutinative condition was easier to acquire for both groups, but the fusional condition was significantly harder for Hungarian speakers. These results support claims of fusional morphology being harder to acquire than agglutinative morphology, especially if learners have no native-language experience with the less “transparent” system.

Our study is a first step toward measuring the effect of diverse factors on the learnability of morphology, and ultimately toward a learning theory that can account for it.



A perceptual magnet effect in pitch accents: parametric modelling 

Joe Rodd, Utrecht University

The presence of a perceptual magnet effect (PME, Kuhl 1991) is strong evidence of category-internal structure, implying that perception and mental representation of the feature in question are compatible with exemplar theory. PME has not previously been conclusively found in pitch accents (linguistically significant fundamental frequency (F0) movements), but has been found in boundary tones, a related intonational event. “Parametric” modelling of speech melody, such as the CoPaSul model (Reichel 2014), involves using mathematical techniques to stylize the movements in the fundamental frequency of speech, extracting parameters describing the shape of the contour. In this investigation,  perceived goodness and discriminability of re-synthesised Dutch L*H rise contours were evaluated by naive participants. The variation between these stimuli was quantified in terms of these polynomial parameters, in place of a more traditional “measurement” approach where, for instance, excursion size, peak alignment and pitch register are used to quantify pitch variation. Using this polynomial-parametric approach to calculate the perceptual distance between different stimuli, PME was detected: (1) rated “goodness” decreased as perceptual distance increased, and (2) equally spaced items far from the prototype were less frequently generalized than equally spaced items in the neighborhood of the prototype.



When getting out of line is in order: acquiring ordinals and ordinality

Caitlin Meyer, University of Amsterdam

Developmental psychologists have repeatedly shown that children discover the exact meanings of cardinals one through four in a slow and stepwise fashion before becoming fully competent counters (i.a. Le Corre & Carey 2007). Linguistic knowledge is argued to play an important role in this process, both in the initial stages and in helping children overcome the limitations of innate, nonverbal, number systems (e.g. Carey 2009, Izard et al. 2008). Something similar applies to ordinals: children need to learn which counting principles to apply and they need linguistic cues and forms to do so. Surprisingly, however, ordinal acquisition and how it relates to children’s understanding of cardinals is generally poorly understood (Colomé & Noël 2012, Koch et al. 2015).

This talk reports on a systematic investigation of how and when Dutch ordinals are acquired and how this relates to cardinal acquisition. I discuss results from two modified Give-a-number tasks (cf. Wynn 1992, among many others) administered to Dutch monolinguals (N=77, ages 2;11–6;4 and N=56, ages 2;8–4;11, respectively). The findings show the tiered acquisition of cardinals also holds for Dutch, but that ordinal acquisition follows a different route, in which linguistic knowledge seems to play a role. I go into children’s response patterns, which together suggest that (i) children may use the singular-plural distinction to determine that ordinals refer to an individual (rather than a set), (ii) ordinals are derived via rules rather than stored lexically and hence, irregular ordinals pose a problem, and (iii) superlative morphosyntax might be involved as well (see also Barbiers 2007). I argue that, though the cardinal acquisition pattern seems to be universal across cultures and languages, the timing and pattern of ordinal acquisition is influenced by language-specific factors.



The acquisition of English recursive NPs: What can we say about the limits in the power of recursion?

Ana T. Pérez-Leroux University of Toronto

Merge (formulated as a recursive function) underlies the cognitive capacity to manipulate symbols in a recursive fashion that is central to human nature/the creative property of language (Hauser, Chomsky & Fitch 2003). Therefore evidence of recursive structures should be present in all human languages (Roberts 2015). Evidence show that recursion in children is very constrained, which is not surprising given the complexity of the semantic contribution of phasal embedding (Arsejinovic & Hinzen 2011). However, it remains problematic that children go through separate steps in their acquisition of recursion, mastering first a single level of embedding of a category (Pérez-Leroux et al 2011), and only later learning to recursively embed the same category type. This raises the question of why children who have learned a rule cannot apply it iteratively. I will present data from a second study eliciting recursively modified NPs in 50 English speaking monolingual children (aged 4;0-5;11). I will discuss data on i) the contrast between recursive and non-recursive modification in children’s production, ii) alternative strategies children employ when producing recursive modification, including evidence of overelaboration and of linearization problems, and iii) on the individual differences between children who produce recursive modification and those who do not.


Arsenijevic, Boban, and Wolfram Hinzen. 2012. On the absence of X-within-X recursion in human grammar. Linguistic Inquiry 43:423–440.

Hauser, Marc D., Noam Chomsky, and W. Tecumseh Fitch. 2002. The faculty of language: What is it, who has it, and how did it evolve? Science 298:1569– 1579.

Pérez-Leroux, Ana Teresa, Anny P. Castilla-Earls, Susana Bejar, and Diane Mas- sam. 2012a. Elmo’s sister’s ball. the development of nominal recursion in children. Language Acquisition 19:301–311.

Roeper, Tom. 2011. The acquisition of recursion: How formalism articulates the child’s path. Biolinguistics 5:57–86.

Roeper, Tom, and William Snyder. 2005. Language learnability and the forms of recursion. In Ug and external systems: Language, brain and computation, ed. Anna Maria DiSciullo, 133–166. Amsterdam: John Benjamins.


Upper-bound numeral modifiers do not impose an upper bound uniformly: Experimental evidence from Greek

Stavroula Alexandropoulou & Sofia Bimpikou (UiL OTS)

This study focuses on the upper-bound numeral modifiers (NMs) mehri (‘up to’) and to poli (‘at most’) in Greek, whose English equivalents share the same semantics and pragmatics according to Nouwen (2010). Specifically, it investigates whether those NMs impose an upper bound to the same extent. Schwarz et al. (2012) via a series of diagnostics comparing up to vs. at most conclude that different semantics have to be assigned to those NMs. Blok (2015) shows that, contrary to at most, the upper bound with up to is not entailed, but implicated, as it can be cancelled (see the contrast in (1)). Blok further argues that this difference can account for the differences Schwarz et al. observe.

(1) Peter is allowed to choose {up to/#at most} ten presents, perhaps even more. 

By means of an experiment, we show that 1) in the case of to poli (‘at most’), the upper bound is derived from its semantics, and 2) in mehri (‘up to’), it is derived as a pragmatic inference. Crucially, our findings lend support to Blok’s account and shed important light on Nouwen’s semantic classification of NMs, which takes the NMs in question to be semantically and pragmatically synonymous.


Cross-domain pitch perception in infancy, an EEG study

Ao Chen, UiL OTS

We examined how pitch perception across different domains develops in infancy. 4- and 12-month-old English infants’ event related potentials (ERP) to musical and speech pitch were recorded. In two conditions, we presented three-note musical pitch and Chinese lexical tones to the infants respectively. The musical pitch and lexical tones had identical duration and shared a similar pitch contour. In each condition, the infants passively listened to 600 trials, with one “standard” stimulus occurring at 80% of the time, and another “deviant” stimulus occurring at 20% of the time. The responses to deviants were compared with the responses to standards to find out how the neural detection of the change developed. For the 4-month-olds, a late positive mismatch response was found in the music condition, and an earlier left-lateralized positive response in the lexical tone condition. For the 12-months, in both conditions, the infants showed an adult-like left-lateralized mismatch negativity (MMN) response. Specifically, the 12-month-olds showed a more pronounced MMN than their adult counterpart in the lexical tone condition. These indicate that the neural responses to complex auditory pitch change mature by 12 months; by 12 months, infants have not fully developed a language specific neural mechanism for pitch change detection.


Incremental interpretation: experimental evidence and an ACT-R analysis

Jakub Dotlacil (joint work with Adrian Brasoveanu)

The main question of the talk is whether meaning representations used in formal semantics are built up incrementally and if so, how incremental interpretation can be modeled in a processing theory.

In the talk, I will first identify phenomena that can tease apart the syntactic and semantic components of interpretation process. This is crucial because the incrementality of meaning composition that is syntax based cannot provide an unambiguous window into the nature of semantic representation building: the incremental nature of real-time compositional interpretation could be due to our processing strategies for building syntactic representations.

I will show that the interaction of presupposition resolution with conjunctions vs. conditionals might provide the right type of evidence. Consider (1) and (2).

(1) Tina will have coffee with Alex again AND she had coffee with him at the local café.

(2) Tina will have coffee with Alex again IF she had coffee with him at the local café.

Both sentences have presupposition in their first part, stating that Tina had coffee with Alex before. This presupposition has to be resolved (i.e., it has to be established in the text that indeed, Tina had coffee with Alex before). For semantic reasons that I’ll explain in the talk AND in (1) signals that the presupposition will not be resolved in the rest of the sentence, while IF in (2) allows for presupposition resolution. This expectation is purely semanticaly driven: there is nothing about the syntax of conjunction vs. if-adjunction that could make the possibility of a successful presupposition resolution more or less likely.

I will discuss two self-paced reading experiments providing evidence that people make use of such semantic information in real time and have expectation about the possibility of presupposition resolution in (2), in contrast to (1). This strongly suggests that their semantic representation is incremental.

The last part of the talk will show how the findings can follow if we embed Discourse Representation Theory (Kamp, 1981) into the cogntive architecture ACT-R.


Statement Processing and Information Believing                                                                              

Myrto Pantazi, ULB

Information we receive through written or oral statements influences our thoughts and actions. It is, thus, important to understand how people endorse or reject this information.
On the one hand, Gilbert et al. (1990, 1993) argued that people are incapable of comprehending statements without initially believing them. For them, believing a statement is part of understanding it, while disbelieving is cognitively and temporarily secondary. To test this hypothesis, they compared the capacity of cognitively depleted and non-depleted individuals to disbelieve statements that were explicitly presented to them as false. They found that cognitively depleted participants remembered more false statements as true, and were more influenced by them in consequent judgments.                                                                                                                                                                                                     On the other hand, Hasson et al. (2005) and Richter et al. (2010) undermined Gilbert et al.’s position, claiming rather that humans possess a routine cognitive mechanism of epistemic vigilance. They both showed that cognitive depletion leads to higher believing only when participants are presented false statements that are not particularly meaningful to them. In the same lines, Sperber et al. (2010) have argued that epistemic vigilance is embedded in human linguistic communication, resulting from a suit of human cognitive mechanisms.                                                                                                    Contrary to this view, and in line with the thesis of Gilbert et al., we present 4 experiments suggesting that humans are inherently ‘gullible’ towards statements they understand. We presented participants with ostensible crime reports containing true and false statements, either uttered by different speakers (Study 1, 3 & 4) or presented in different font colors (Study 2). In Studies 1 & 2 we depleted half of the participants by a simultaneous task (counting the occurrence of a specific verb in the reports, and detecting a digit on the screen, respectively). We detected a gullibility bias for both depleted and non-depleted participants, namely a tendency to remember more false statements as true than true statements as false, and an impact of false statements on judgments they made about the crime’s perpetrators.                                                          In Study 3, based on Tetlock (1983), we tried to increase participants’ vigilance by informing them that they will have to orally account for their judgments of the perpetrators to the experimenters. Still, accountable participants displayed the above-mentioned gullibility bias. Finally, in Study 4 the gullibility bias persisted even when we informed participants that the false statements in the reports were uttered by a lawyer aiming either at incriminating or exculpating the perpetrator. Overall, our results corroborate the view that people strongly tend to believe statements they understand. Crucially, this tendency appears stronger than previously reported, i.e. in the absence of cognitive depletion, and in contexts where people should be able to process information in a vigilant way. 


Post-focus in Shanghai Chinese

Lei Sun, Leiden University

        It is well-known that effective speech communication does not only depend on what the speaker is saying, but also the way how he is saying. Take (1) & (2) for example. In (1), the speaker wants to emphasize the fruit Mary bought is  an apple not an orange, “apple” would be produced with more prominence. In (2), if the speaker needs to testify that Mary did not steal an apple, but she bought one, the speaker would produce the verb “bought” with more prominence while “apple” in this case would be produced with less prominence as given information. The way how the speaker packages information differently is called information structure. Focus is one of information structure notions. In the examples, to correct wrong information, the speaker produces focused words with high prominence and the same words with less prominence at the post-focus position and such a way is commonly recognized as prosodic encoding of information structure notions.

(1)    Mary bought an orange in the supermarket. 

         Mary bought an APPLE in the supermarket. 

(2)    Mary stole an apple in the supermarket. 

         Mary BOUGHT an apple in the supermarket.

Much work has been done on how information structure notions are prosodically encoded in languages (Baumann, Grice & Steindamm, 2006; Ishihara & Fery, 2006; Xu, 1997; Chen, 2010). In Chinese languages, pitch range manipulation has been long considered an important strategy of realizing focus: focus expands the pitch range of focused constituents and suppresses the pitch range of post-focused constituents while keeping that of pre-focus words largely intact (Xu, 1999). However, in recent years, a few studies have shown that pitch range compression is not a reliable cue for tonal realization in the post-focus condition as pitch range expansion is also found in some cases. Therefore, an alternative view holds that both post-focus compression and expansion belong to the multifaceted realizations of the weak implementation of post-focus tonal targets, which results from their non-prominent position in prosodic structure (Chen, 2010). To test this hypothesis, an experiment was done to investigate how prosodic realizations of post-focus constituents are constrained by prosodic structure in Shanghai Chinese, a tonal language which is not intelligible with Standard Chinese. We have found that in this language post-focus f0 realizations were absent at the level of Prosodic Word and at the higher level of Phonological Phase pitch range was expanded in some cases and only pitch register lowering was consistently found. We tentatively concluded that pitch register lowering is a phrasal marker in Shanghai Chinese and post-focus realization is constrained by prosodic structure.


Perception of lexical pitch-accent by Korean learners of Japanese

Shuangshuang Hu, Utrecht University

Learners of foreign languages have difficulties in processing the target language in many aspects. Phonology is one of these. Previous studies have shown that second language (L2) learners encounter difficulties in parsing segmental (e.g., vowels and consonants) and prosodic features (e.g., stress and tone). For instance, difficulty in discriminating /r/ vs. /l/ is persistent in Japanese learners of English [1]. Advanced French learners of Spanish have ‘persistent deafness’ in perceiving word stress [2]. Theoretical models have been proposed to account for such perceptual difficulties, in particular why certain phonological contrasts in the native language influence non-native perception [3-4]. In order to examine the predictions from these models, I investigated L2 perception of lexical pitch-accent. Two language groups were tested: advanced Korean learners of Japanese (target group) and Tokyo Japanese native speakers (control group). Tokyo Japanese has a lexical pitch-accent system where the meaning of a word is determined by the presence or location of a pitch fall (or “pitch-accent”) in multi-syllable words. In contrast, Seoul Korean features no word-level tonal representations to signal lexical contrasts.  Three accentual contrasts (H’-L vs. L-H’, H’-L vs. L-H and L-H vs. L-H’) were investigated by means of a sequence-recall experiment adapted from Dupoux et al. (2001). The results showed that advanced Korean learners performed as well as native Japanese speakers in perceiving pitch-accent contrasts. This result cannot be explained by the models in [3-4] but instead supports models predicting that non-native perception of certain cues can be enhanced or reduced, depending on their functional value in the native language [5-7]. Although pitch does not function at word-level in Seoul Korean, it plays a crucial role in word segmentation [8]. In particular, IP (intonational phrases) and AP (accentual phrases) depend to a large extent on the usage of pitch to distinguish phrase-level meanings in Seoul Korean. This provides some insight into Korean learners’ perception of Japanese pitch-accent in that the usage of pitch at the phrasal level in their native language facilitates the perception of lexical pitch in pitch-accent.

[1] Guion, S. G., Flege, J. E., Akahane-Yamada, R.,  Pruitt, J. C. 2000. An investigation of current models of second language speech perception: The case of Japanese adults’ perception of English consonants. Journal of the Acoustical Society of America, 107, 2711–2724

[2] Dupoux, Sebstian-Galles, E., N., Peperkamp, S. 2008. Persistent stress “deafness”: the case of French learners of Spanish. Cognition 106: 682-706.

[3] Lado, R. 1957. Linguistics across cultures: Applied linguistics for language teachers. Ann Arbor, MI: University of Michigan Press.

[4] Brown, C. 1998. The role of the L1 grammar in the L2 acquisition of segmental structure. Second Language Research, 14, 136–193.

[5] Francis, A. L., Nusbaum, H. C. 2002. Selective attention and the acquisition of new phonetic categories. Journal of Experiment Psychology: Human Perception and Performance, 28, 349–366.

[6] Jusczyk, P. W. 1997. The discovery of spoken language. Cambridge, MA: MIT Press.

[7] Iverson, P., Kuhl, P. K., Akahane-Yamada, R., Diesch, E., Tohkura, Y., Kettermann, A., et al. (2003). A perceptual interference account of acquisition difficulties for non-native phonemes. Cognition, 87, B47–B57.

[8] Kim, S., Cho, T. 2009. The use of phrase-level prosodic information in lexical segmentation: Evidence from word-spotting experiments in Korean. Journal of the Acoustical Society of America, 125 (5), 3373-3386.


On the interpretation of intonation

Stella Gryllia

Leiden University Centre for Linguistics

In this talk I will discuss two cases where intonation is relevant for interpretation. The first case concerns the pragmatics of two melodies commonly used with Greek wh-questions, L*H L-!H%, described as the default, and LH* L-L% considered less frequent and polite. Two hypotheses were tested (a) the !H%-ending melody is associated with information-seeking questions, while the L%-ending melody is pragmatically more flexible and thus appropriate also for non-information-seeking wh-questions expressing bias; (b) the !H%-ending melody, being more polite, is more appropriate for female talkers, all else being equal. In Experiment 1, comprehenders rated !H ending and L%-ending versions of the same questions for politeness and appropriateness for the context in which they were heard (which favored either information-seeking or “biased” wh-questions). In Experiment 2, comprehenders heard the same questions and chose between two follow-up responses, one providing information, the other addressing the bias of the wh-question. Comprehenders rated !H%-ending questions more appropriate than L%-ending questions and judged the !H%-ending questions of female talkers more polite. They also chose information-providing answers more frequently after !H%- than L%-ending questions, but the preference was higher for female talkers and depended on comprehender gender. The results argue in favor of a compositional view of intonational meaning which depends not only on the tune but also on context, broadly construed.

The second case concerns question bias in questions with negative polarity items. The experiment tested two hypotheses: (a) the emphatic NPI kanenas ‘no-one/nobody’ in a negative polar question creates a bias for a negative answer; (b) the non-emphatic NPI kanenas ‘anyone/anybody’ creates no specific bias for an answer. In the experiment, comprehenders heard three types of questions (i. a positive polar question including a non-emphatic NP1, ii. a negative polar question including a non-emphatic NPI, iii. a negative polar question including an emphatic NPI) and chose between two follow-up responses, a positive answer and a negative one. Both hypotheses were verified.


Individual differences in language comprehension: a behavioral study

Carolien van den Hazelkamp & Rachel Gargiulo (Utrecht University)

Traditionally, psycholinguistic research is built on the implicit assumption that the process of understanding language does not differ between individuals. However, a number of studies have shown that this assumption may not be valid: while individual differences are not visible using traditional analysis techniques that average over all participants, analysis on the individual level reveals that these differences do indeed exist (Tanner & Van Hell 2014). Multi-stream models of language comprehension such as the one put forward by Kuperberg (2007) provide a useful framework to look at variation between individuals. We hypothesize that the nature of this variation lies in the degree of reliance on or attention paid to different cues in linguistic input – roughly summarized as semantic versus syntactic. In other words, we hypothesize that individuals each have a dominant processing stream.

In a series of three behavioral experiments (a self paced reading experiment, an auditory violation detection experiment and an experiment based on a Speed-Accuracy Tradeoff paradigm (cf. McElree)) we aimed to test this hypothesis and explore the nature and possible sources of individual variation. To be able to look into other possible sources of individual variability, we took background measures of handedness, familial sinistrality, working memory and executive control.

The data of our experiments shows that there are indeed differences between individuals in performance on the different experiments, and that these differences are not arbitrary: performance on one experiment is related to performance in the other experiments, albeit that this relation is not the one that we predicted on the basis of our hypothesis.

Tanner, D. & van Hell, J. G. (2014). ERPs reveal individual differences in morphosyntactic processing. Neuropsychologia, 56, 289-301.

Kuperberg, G. R. (2007). Neural mechanisms of language comprehension: Challenges to syntax. Brain Research, 1146, 23–49.



Differences and similarities between scalar inferences and scalar modifiers

Yaron McNabb (Utrecht University)

Speakers that utter a sentence like in (1) are usually understood to communicated that they didn’t mean that all of Ann’s friends shared her grim view of the world. In oder to explain this scalar implicature, the scalar expression a handful is assumed to evoke a set of alternatives whose members are ordered in terms of informativeness along the lines of (2). A listener reasons that the speaker used the most informative expression, and so informationally-stronger alternatives like most or all don’t hold (Horn 1972, Gazdar 1979, Levinson 2000, Geurts 2010,inter alia).


(1) A handful of Ann’s friends shared her view of the world as a dark and dangerous place.

(2) <some, a handful, a bunch, many, most, all>


Many important questions have been addressed regarding scalar implicature, such as the processing of such inferences (e.g. Bott & Noveck 2004, Noveck & Posada 2003, Huang & Snedeker 2009) and whether they are part of a grammaticalized mechanism rather than pragmatic inferences (Chierchia, 2006; Fox, 2007, inter alia). What is striking about many of these studies is that the empirical investigation was confined mostly to some vs. all, neglecting the cross-categorial pervasiveness of the phenomena, as observed by, e.g., Horn (1972) and Hirschberg (1985). The few studies that looked at a more diverse group of scalar implicatures have found that different scalar expressions differ in the extent to which they give rise to scalar implicature (Doral et al. 2009, 2012, van Tiel et. al ms.).


In this talk, I present results from an experimental investigation of the extent to which various scalar expressions give rise to scalar implicature. I look at quantity expressions like a handful, adjectives like large, ranked ordering like lieutenant general and non-scalar expressions like novella. In addition, I examine the semantic effect of modification by the superlative modifiers at least and at most. Modification by at least, as in (3a), is assumed to block the implicature that not all of Ann’s friends have a gloomy view of the world, whereasat most, as in (3b), would impose an upper-bound construal (Krifka 1999, Geurts & Nouwen 2007). The same pattern was investigated for other scalar expressions, such as temporary in (4).


(3) a. At least handful of Ann’s friends shared her view of the world as a dark and dangerous place.

b. At most handful of Ann’s friends shared her view of the world as a dark and dangerous place.

(4) a. They found shelter for hundreds of people on a temporary basis.

b. They found shelter for hundreds of people on at least a temporary basis.

c. They found shelter for hundreds of people on at most a temporary basis.


The results show that stronger alternatives (e.g. all when saying a handful) were more acceptable in at most a handful than in a handful. However, the stronger alternatives were more acceptable in the non-modified case, as in (1) and (4a), for adjectives, as temporary in (4), and ranked orderings and some non-scalars. The results from this study provide further support to findings from Doran et al 2009, 2012 and van Tiel 2012, that not all scalar inferences are the same, and show that the comparison with at most is a successful and informative diagnosis in distinguishing between semantic and pragmatic upper-bound construal. Results from modification by at least reveal that stronger alternatives, e.g. all in (3a), are acceptable but degraded, suggesting differences between ignorance inferences and scalar implicatures.



Prosodic and lexicosyntactic cues in turn-projection by Dutch and English toddlers

Imme Lammertink (Radbout University, Nijmegen)

Successful speech coordination during conversation requires adult speakers to predict upcoming turn transitions with lexicosyntactic, prosodic, and non-verbal information [6]. Infants process prosodic information from early on for speech segmentation [4] and emotion recognition [3], but only learn about lexicosyntactic rules between their second and third year of life [2]. Experimental work focusing on the use of these linguistic cues in turn projection has shown that toddlers use both lexicosyntactic and prosodic cues, but the relative weight of these cues is still unresolved [1,5]. Prior work used phonetic manipulation, such as filtered “underwater” speech, to control the availability of linguistic cues. In the current study, we pit prosodic cues against lexicosyntactic ones in unfiltered, and thus natural-sounding, conversation to test the relative weighing of both cues. To examine the generality of our findings, we tested two linguistic populations: Dutch (experiment 1) and English (experiment 2). We tracked the anticipatory eye-movements of 21 Dutch and 20 English two-year-olds, and 16 Dutch and 20 English adult controls as they watched videos of dyadic puppet conversation. Each target utterance was controlled for lexicosyntactic and prosodic (intonational cues) to turn completion and interrogativity, resulting in four types of targets (fully incomplete, incomplete syntax, incomplete prosody, and fully complete). Cues conflicted in two conditions (incomplete syntax and incomplete prosody) to test for their relative primacy. We found that Dutch and English toddlers and adults used both lexicosyntactic and prosodic cues in their anticipation of upcoming speaker changes. But, when the cues are pitted against each other (incomplete syntax and incomplete prosody), listeners weigh lexicosyntactic cues over prosodic ones. We found no overall differences in the use of the cues for turn-projection between the two languages. The results raise new questions regarding the interaction between lexicosyntactic and prosodic cues in children’s acquisition of conversational predictive processing.


[1] Casillas, M. & Frank, M.C. (under review). Tracking and predicting turn structure during language acquisition.

[2] Clark, E. V. (2003). First Language Acquisition. Cambridge: Cambridge University Press.

[3] Grossmann, T., Striano, T., & Friederici, A., D. (2005). Infants’ electric brain responses to emotional prosody. NeuroReport 16, 1825–1828

[4] Jusczyk, P. W. (1997). The discovery of spoken language. Cambridge, MA: MIT press.

[5] Keitel, A., Prinz, W., Friederici, A., D., Hofsten, v., C., & Daum, M., M. (2013). Perception of conversations: The importance of semantics and intonation in children’s development.  Journal of Experimental Child Psychology 116, 264–277

[6] Levinson, S. C. (2013). “Action formation and ascription” in T. Stivers & J. Sidnell (Eds.), The Handbook of Conversation Analysis (pp. 103–130). Malden, MA: Wiley-Blackwell.



Prosodic and lexicosyntactic cues in turn-projection by Dutch and English toddlers
Imme Lammertink (Radbout University, Nijmegen)

Successful speech coordination during conversation requires adult speakers to predict upcoming turn transitions with lexicosyntactic, prosodic, and non-verbal information [6]. Infants process prosodic information from early on for speech segmentation [4] and emotion recognition [3], but only learn about lexicosyntactic rules between their second and third year of life [2]. Experimental work focusing on the use of these linguistic cues in turn projection has shown that toddlers use both lexicosyntactic and prosodic cues, but the relative weight of these cues is still unresolved [1,5]. Prior work used phonetic manipulation, such as filtered “underwater” speech, to control the availability of linguistic cues. In the current study, we pit prosodic cues against lexicosyntactic ones in unfiltered, and thus natural-sounding, conversation to test the relative weighing of both cues. To examine the generality of our findings, we tested two linguistic populations: Dutch (experiment 1) and English (experiment 2). We tracked the anticipatory eye-movements of 21 Dutch and 20 English two-year-olds, and 16 Dutch and 20 English adult controls as they watched videos of dyadic puppet conversation. Each target utterance was controlled for lexicosyntactic and prosodic (intonational cues) to turn completion and interrogativity, resulting in four types of targets (fully incomplete, incomplete syntax, incomplete prosody, and fully complete). Cues conflicted in two conditions (incomplete syntax and incomplete prosody) to test for their relative primacy. We found that Dutch and English toddlers and adults used both lexicosyntactic and prosodic cues in their anticipation of upcoming speaker changes. But, when the cues are pitted against each other (incomplete syntax and incomplete prosody), listeners weigh lexicosyntactic cues over prosodic ones. We found no overall differences in the use of the cues for turn-projection between the two languages. The results raise new questions regarding the interaction between lexicosyntactic and prosodic cues in children’s acquisition of conversational predictive processing.

[1] Casillas, M. & Frank, M.C. (under review). Tracking and predicting turn structure during language acquisition.
[2] Clark, E. V. (2003). First Language Acquisition. Cambridge: Cambridge University Press.
[3] Grossmann, T., Striano, T., & Friederici, A., D. (2005). Infants’ electric brain responses to emotional prosody. NeuroReport 16, 1825–1828
[4] Jusczyk, P. W. (1997). The discovery of spoken language. Cambridge, MA: MIT press.
[5] Keitel, A., Prinz, W., Friederici, A., D., Hofsten, v., C., & Daum, M., M. (2013). Perception of conversations: The importance of semantics and intonation in children’s development. Journal of Experimental Child Psychology 116, 264–277
[6] Levinson, S. C. (2013). “Action formation and ascription” in T. Stivers & J. Sidnell (Eds.), The Handbook of Conversation Analysis (pp. 103–130). Malden, MA: Wiley-Blackwell.



The wisest or the nicest? The influence of speaker certainty and affect in word learning
Myrthe Bergstra (UiL OTS)

The study aimed to find out whether speaker certainty of the interlocutor and affect towards the interlocutor could guide children in the word learning process. That is, do children prefer to learn new words from speakers that are certain of what they say rather than from speakers that are uncertain? And do children prefer to learn new words from kind speakers rather than from unkind speakers? Which of these factors is guiding the child more?

In three experiments with 4- to 6-year old children, two handpuppets labeled novel objects. It was found that children tend to learn new words from a certain speaker rather than from an uncertain speaker, but they had no significant preference for learning from a kind speaker. When both cues were combined it was found that children prefer an unkind an certain speaker over a kind and uncertain speaker, showing that speaker certainty leads to stronger learning preferences than affect.

However, kind and certain speakers were preferred most, indicating that affect has a small additive effect. To address worries in the field that experiments with set-ups like these do not really address word learning, a posttest was conducted for each experiment in which children had to recall which object belonged to which label. It was found that many of the new links between labels and objects were enduring.



Early prediction of dyslexia: Machine learning in infant data
Ao Chen (UiL OTS)

Dyslexia is a developmental disorder specific to reading and writing, which affects 4-5% of school-age children. Precursors of dyslexia can be found at an early age before the children start learning to read and write (e.g. Koster, Been, Krikhaar, Zwarts, Diepstra, & van Leeuwen, 2005; Kerkhoff, de Bree, de Klerk, & Wijnen, 2013). Although many studies show differences in early speech/language development between familial risk (FR) and typically developing (TD) children, none of these markers have been successfully used to predict later language/literacy performance at individual level. The PODIUM study uses machine learning (ML) techniques to explore if data on speech / language abilities in infancy can predict the occurrence of later literacy difficulties in individual children.

So far, we have used the support vector machine (SVM) to classify FR and TD at 18-19 months. The SVM is a supervised learning algorithm for pattern recognition. In our case the problem consists of separating FR and TD based on infants’ language profiles. The model was trained and tested (using leave-one-out cross-validation) on infants’ vocabulary data, which was collected through a parental survey. The parents filled in the Dutch version of MacArthur-Bates Communicative Development Inventories (NCDI); they were asked to indicate, in a list of 22 semantic categories, the words that their child understood. We trained the model with both the raw counts of words in each semantic category of each infant, and aggregated semantic composites such as common nouns, predicates, and closed class words (Koster et al., 2005). Our preliminary results show that: 1) When using the counts of each semantic category as features, the model yields a cross-validation accuracy of 71%. 2) When the model was trained with aggregated composites, it yielded an accuracy of 65%. Interestingly, the model trained with individual categories indicate that within the composites, the different categories have opposite contributions to the discrimination between TD and FR, which suggests that combining these categories into one composite may partly cancelling out the discrimination effect. We regrouped the individual categories roughly based on the contribution of individual categories, and the prediction accuracy was improved to 78% at individual level. Our results suggest that the previous assumed semantic composites do not fully reflect the differences in vocabulary acquisition between TD and FR, and a finer grained division of children’s vocabulary composites is needed.