Paul Boersma

University of Amsterdam | UVA · Amsterdam Center for Language and Communication

Pitch tracking algorithms can show upward or downward jumps in F0 by one octave. These “octave jumps” are sometimes thought of as pitch-tracking “errors”, in the sense that they constitute a “mistake” in the algorithm. Using Praat software, we discuss the point (which has been made before) that measured octave jumps often actually reflect genuine c...

Shanghai Chinese has a complicated tone system conditioned by phonological onset voicing and syllable structure. Checked tones in Shanghai Chinese are acoustically realized with tenser phonation, shorter duration and a more central vowel space than their unchecked counterparts. The present study examines whether these cues also play a role in speec...

--A more in-depth discussion of the same topic can be seen in our paper for Interspeech 2024-- Pitch doubling and halving occur when the F0 pitch contour does not follow a consistent curve but rather ‘jumps’ in pitch are observed. Often these pitch jumps are deemed as pitch tracking ‘errors’, due to the initial seemingly inconsistent nature of the...

Introduction Research indicates that statistical learning plays a role in word learning by enabling the learner to track the co-occurrences between words and their visual referents, a process that is named cross-situational word learning. Word learning is problematic for children with developmental language disorder (DLD), and a deficit in statisti...

Theories of phonology claim variously that phonological elements are either innate or emergent, and either substance-full or substance-free. A hitherto underdeveloped source of evidence for choosing between the four possible combinations of these claims lies in showing precisely how a child can acquire phonological elements. This article presents c...

Listeners are sensitive to speech sounds’ probability distributions. Distributional training (DT) studies with adults typically involve conscious activation of phoneme labels. We show that distributional exposure can shift existing phoneme boundaries (Spanish /e/–/i/) pre-attentively. Using a DT paradigm involving two bimodal distributions we asses...

Categorization of sensory stimuli is a vital process in understanding the world. In this paper we show that distributional learning plays a role in learning novel object categories in school-aged children. An 11-step continuum was constructed based on two novel animate objects by morphing one object into the other in 11 equal steps. Forty-nine chil...

Since Saffran, Aslin and Newport (1996) showed that infants were sensitive to transitional probabilities between syllables after being exposed to a few minutes of fluent speech, there has been ample research on statistical learning. Word segmentation studies usually test learning by making use of “offline methods” such as forced-choice tasks. Howev...

Several studies have signaled grammatical difficulties in individuals with developmental dyslexia. These difficulties may stem from a phonological deficit, but may alternatively be explained through a domain-general deficit in statistical learning. This study investigates grammar in children with and without dyslexia, and whether phonological memor...

Using an individual differences approach in children with and without dyslexia, this study investigated the hypothesized relationship between statistical learning ability and literacy (reading and spelling) skills. We examined the clinical relevance of statistical learning (serial reaction time and visual statistical learning tasks) by controlling...

This paper argues that if phonological and phonetic phenomena found in language data and in experimental data all have to be accounted for within a single framework, then that framework will have to be based on neural networks. We introduce an artificial neural network model that can handle stochastic processing in production and comprehension. Wit...

Children with Developmental Language Disorder (DLD) have difficulties acquiring the grammatical rules of their native language. It has been proposed that children’s detection of sequential statistical patterns correlates with grammatical proficiency and hence that a deficit in the detection of these regularities may underlie the difficulties with g...

Visual statistical learning (VSL) has been proposed to underlie literacy development in typically developing (TD) children. A deficit in VSL may thus contribute to the observed problems with written language in children with dyslexia. Interestingly, although many children with developmental language disorder (DLD) exhibit problems with written lang...

Human-robot interaction (HRI) research aims to design natural interactions between humans and robots. Intonation, a social signaling function in human speech investigated thoroughly in linguistics, has not yet been studied in HRI. This study investigates the effect of robot speech intonation in four conditions (no intonation, focus intonation, end-...

Visual statistical learning (VSL) was traditionally tested through offline two-alternative forced choice (2-AFC) questions. More recently, online reaction time (RT) measures and alternative offline question types have been developed to further investigate learning during exposure and more adequately assess individual differences in adults (Siegelma...

Statistical learning (SL) difficulties have been suggested to contribute to the linguistic and non-linguistic problems observed in children with dyslexia. Indeed, studies have demonstrated that children with dyslexia experience problems with SL, but the extent of the problems is unclear. We aimed to examine the performance of children with and with...

The purpose of this study is to describe the acoustic changes of the consonants /t/ and /d/ in Dutch speaking individuals before and after total laryngectomy. The speech of seventeen participants was recorded before and after treatment. Eighteen tokens from a read-aloud text were obtained with /t/ or /d/ in initial position. Prevoicing, burst durat...

Successful language use requires the ability to process nonadjacent dependencies (NADs) that occur in linguistic input. Learning such structural regularities seems therefore crucial for children, and researchers have indeed proposed that language problems in children with developmental language disorder (DLD), especially problems with grammar, are...

While learning language, children unconsciously detect and process structural regularities that reflect (morpho)syntactic rules of language. This mechanism is referred to as statistical learning (SL). A link between SL and children’s knowledge of (morpho)syntactic rules (i.e. grammar) has been proposed. While intuitively such a link may be evident,...

The present study investigated the hypothesized relationship between statistical learning (SL) ability and literacy skills by adopting an individual differences approach in children with and without dyslexia. We analyzed both reading and spelling performance and used two SL measures that have previously been linked to literacy attainment (serial re...

Nonadjacent dependency learning is thought to be a fundamental skill for syntax acquisition and often assessed via an offline grammaticality judgment measure. Asking judgments of children is problematic, and an offline task is suboptimal as it reflects only the outcome of the learning process, disregarding information on the learning trajectory. Th...

In this poster, we compared the performance of 50 children with dyslexia (26 girls, 24 boys, aged 8;4 - 11;2) with 50 age- and gender-matched children without dyslexia on three distinct tasks that assessed their statistical learning abilities: a serial reaction time, visual statistical learning and auditory nonadjacent dependency learning task.

Cambridge Core - Research Methods in Linguistics - edited by Robert J. Podesva

Research Methods in Linguistics - edited by Robert J. Podesva January 2014

Purpose The current meta-analysis provides a quantitative overview of published and unpublished studies on statistical learning in the auditory verbal domain in people with and without specific language impairment (SLI). The database used for the meta-analysis is accessible online and open to updates (Community-Augmented Meta-Analysis), which facil...

The aim of this paper is to show that a sequence of typologically not unusual sound changes has led to three conspicuous properties of the dialects in a large connected area of Low and Central Franconian. First, these dialects have a binary contrast between acute and circumflex tones. Second, the majority of these dialects (“group A”) show length r...

Background: Literacy impairments in dyslexia have been hypothesized to be (partly) due to an implicit learning deficit. However, studies of implicit visual artificial grammar learning (AGL) have often yielded null results. Aims: The aim of this study is to weigh the evidence collected thus far by performing a meta-analysis of studies on implicit...

This poster describes preliminary results on novel online measures of nonadjacent dependency learning in children with and without specific language acquisition. Data collection is still ongoing.

In multilevel parallel Optimality Theory grammars, the number of candidates (possible paths from the input to the output level) increases exponentially with the number of levels of representation. The problem with this is that with the customary strategy of listing all candidates in a tableau, the computation time for evaluation (i.e., choosing the...

Visual statistical learning (VSL) is usually tested through offline two-alternative forced choice (2-AFC) questions, which has yielded mixed results in children. We assessed children’s VSL using an online reaction time (RT) measure and two distinct offline question types to investigate whether these methods can track learning over time and outcome...

Results of our pilot with 46 Typically developing children showing that reaction times can be used as measure of online nonadjacent dependency learning

Correlational study investigating the link between implicit learning and spelling ability in adults

In their first year, infants' perceptual abilities zoom in on only those speech sound contrasts that are relevant for their language. Infants' lexicons do not yet contain sufficient minimal pairs to explain this phonetic categorization process. Therefore, researchers suggested a bottom-up learning mechanism: infants create categories aligned with t...

Infants’ perception of speech sound contrasts is modulated by their language environment, for example by the statistical distributions of the speech sounds they hear. Infants learn to discriminate speech sounds better when their input contains a two-peaked frequency distribution of those speech sounds than when their input contains a one-peaked fre...

Distributional learning of speech sounds is learning from simply being exposed to frequency distributions of speech sounds in one’s surroundings. In laboratory settings, the mechanism has been reported to be discernible already after a few minutes of exposure, in both infants and adults. These “effects of distributional training” have traditionally...

Distributional vowel training for adults has been reported as " effective " for Spanish and Bulgarian learners of Dutch vowels, in studies using a behavioural task. A recent study did not yield a similar clear learning effect for Dutch learners of the English vowel contrast /ae/~/ε/, as measured with event-related potentials (ERPs). The present stu...

Distributional learning of speech sounds (i.e., learning from simple exposure to frequency distributions of speech sounds in the environment) has been observed in the lab repeatedly in both infants and adults. The current study is the first attempt to examine whether the capacity for using the mechanism is different in adults than in infants. To th...

An important mechanism for learning speech sounds in the first year of life is “distributional learning,” i.e., learning by simply listening to the frequency distributions of the speech sounds in the environment. In the lab, fast distributional learning has been reported for infants in the second half of the first year; the present study examined w...

In previous research on distributional training of non-native speech sounds, distributions were always discontinuous: typically, each of only eight different stimuli was repeated multiple times. The current study examines distributional training with continuous distributions, in which all presented tokens are acoustically different. Adult Spanish l...

We present a method for assessing categorical perception from continuous discrimination data. Until recently, categorical perception of speech has exclusively been measured by discrimination and identification experiments with a small number of different stimuli, each of which is presented multiple times. Experiments by Rogers and Davis (2009), how...

Certain kinds of sound inventories are more widely found in the world's languages than other kinds. Identifying a language with its speakers, we can state that languages show a preference for certain kinds of sound systems at the cost of other ones. Every language has attained its sound inventory in the course of a long-termed development involving...

In Croatia different folk singing styles have been inherited over long periods of time. To examine the differences between the klapa and dozivački styles of singing, singing voices of 12 professional male singers were digitally recorded and analysed for long-term average spectrum (LTAS) in the Praat program. All the singers were members of the LADO...

This paper addresses remarks made by Flemming (2003) to the effect that his analysis of the interaction between retroflexion and vowel backness is superior to that of Hamann (2003b). While Hamann maintained that retroflex articulations are always back, Flemming adduces phonological as well as phonetic evidence to prove that retroflex consonants can...

This paper presents an Optimality-Theoretic (OT) grammar model that is intended to be capable of handling ‘all ’ of phonology: its representations with their relations, its processes with their relations, its connection to the semantics, its acquisition by the child, its evolution over the generations, and its typology across languages. The goal of...

This paper examines four acoustic properties (duration F0, F1, and F2) of the monophthongal vowels of Iberian Spanish (IS) from Madrid and Peruvian Spanish (PS) from Lima in various consonantal contexts (/s/, /f/, /t/, /p/, and /k/) and in various phrasal contexts (in isolated words and sentence-internally). Acoustic measurements on 39 speakers, ba...

It has been observed that in production, the boundary between the vowels /i/ and /e/ is diagonal, i.e. it involves both F1 and F2; in perception, by contrast, the boundary has been observed to be horizontal, i.e. listeners do not use F2 as a cue for distinguishing the two vowels. The same is true of the /u–o/ boundary. With computer simulations of...

We present a method for assessing categorical perception from continuous discrimination data. Until recently, categorical perception of speech has exclusively been measured by discrimination and identification experiments with a small number of repeatedly presented stimuli. Experiments by Rogers and Davis [1] have shown that using non-repeating sti...

Cross-language perception provides insight into the use of perceptual cues to native segments and their application to segments in a different language. In the present study we test the perception of the three Dutch labiodentals /f, v, ʋ/ by listeners of German, English, Croatian and Polish in a forced-choice identification task. We test whether th...

For many different reasons, speakers borrow words from other languages to fill gaps in their own lexical inventory. The past ten years have been characterized by a great interest among phonologists in the issue of how the nativization of loanwords occurs. The general feeling is that loanword nativization provides a direct window for observing how a...

This article shows that Error-Driven Constraint Demotion (EDCD), an error-driven learning algorithm proposed by Tesar (1995) for Prince and Smolensky’s (1993/2004) version of Optimality Theory, can fail to converge to a correct totally ranked hierarchy of constraints, unlike the earlier non-error-driven learning algorithms proposed by Tesar and S...

This paper examines four acoustic correlates of vowel identity in Brazilian Portuguese (BP) and European Portuguese (EP): first formant (F1), second formant (F2), duration, and fundamental frequency (F0). Both varieties of Portuguese display some cross-linguistically common phenomena: vowel-intrinsic duration, vowel-intrinsic pitch, gender-dependen...

Sorry, there is no abstract.

We investigate whether there is a within-speaker effect of a higher F0 on the values of the first and the second formant. When asked to speak at a high F0, speakers turn out to raise their formants as well. In the F1 dimension this effect is greater for women than for men. We conclude that while a general formant raising effect might be due to the...

We present a simple algorithm for running a listening experiment aimed at finding the best exemplar in a multidimensional space. For simulated humanlike listeners, who have perception thresholds and some decision noise on their responses, the algorithm on average ends up twelve times closer than Iverson and Evans’ goodness interpolation algorithm.

This paper shows how one can formalize the phonology-phonetics interface within constraint-based frameworks such as Optimality Theory (OT) or Harmonic Grammar (HG) and why it is necessary and advantageous to do so. I will describe the phonology-phonetics interface in terms of cue constraints and illustrate their workings in interaction with each ot...

This paper aims to provide acoustic description and comparison of vowel properties of two major dialects of Spanish: Peruvian and European. The method of data collection and acoustic analyses reported in Escudero et al. (submitted) for Portuguese is replicated in order to allow for future comparisons across languages. Each of the five Spanish monop...

This paper reconciles the standpoint that language users do not aim at improving their sound systems with the observation that languages seem to improve their sound systems. Computer simulations of sibilant inventories show that constraint-based learners who optimize their perception grammars automatically introduce a so-called prototype effect, i....

This article shows that the usual speaker-based account of h-aspiré in French can explain at most three of the four phonological processes in which it is involved, whereas a listener-oriented account can explain all of them. On a descriptive level, the behaviour of h-aspiré is accounted for with a grammar model that involves a control loop, whose c...

abstract phonological surface form from raw auditory material !Speech perception is constrained by the familiar language-specific structural constraints , phonologicalp erception 4 Bidirectionalt wo-levelO Tm odels of

We investigate how supervision (in the form of explicit instruction) interacts with distributional learning in the acquisition of the perception of a novel vowel contrast in a second language. An experiment with non-Dutch-speaking Bulgarians reveals that listeners who receive bimodal distributional training without explicit instruction can acquire...

In recent work (Boersma & Hayes 2001), Stochastic Optimality Theory has been used to model grammaticality judgments in exactly the same way as corpus frequencies are modelled, namely as the result of noisy evaluation of constraints ranked along a continuous scale. It has been observed, however, that grammaticality judgments do not necessarily refle...

Optimality Theory has met with a bad press in the more emergentist (e.g. computational) literature for its reliance on innate constraints and even on innate constraint rankings (positional faithfulness, licensing by cue). In this talk I will show with computer simulations that even if the learner's initial grammar starts with a large number of cons...

This paper examines the differences between three Croatian folk singing styles, namely klapa, ojkanje, and tarankanje. In order to factor out singer-specific properties, each of the styles was performed by the same 12 professional male singers. The 36 performances were analyzed with a long-term average spectrum (LTAS) method from which direct effec...

1. Bidirectional phonology and phonetics 1a. A single grammar for phonology and phonetics: four representations Underlying Form Surface Form Auditory Form Articulatory Form lexical constraints faithfulness constraints structural constraints cue constraints sensorimotor constraints articulatory constraints { { phonological representations phonetic r...

The task of the listener: comprehension ‘meaning’ Šunderlying formŠ /surface form/ [auditory form] [articulatory form] (Boersma 2005, Apoussidou 2006) The task of the speaker: production – 2 – The point: markedness can go In this talk I will first illustrate that phonologists have traditionally invoked the concept of ‘markedness ’ to explain a vari...

Morrison (2005) criticizes the analytical and statistical methods that Escudero & Boersma (2004), henceforth E&B, used for assessing the cue weightings of the participants in their listening experiments. He proposes that logistic regression constitutes a better method for measuring perceptual cue weighting than E&B's'edge difference ratio'. The pre...

This article shows that the usual speaker-based account of h-aspir茅 in French can explain at most three of the four phonological processes in which it is involved, whereas a listener-oriented account can explain all of them. On a descriptive level, the behaviour of h-aspir茅 is accounted for with a grammar model that involves a control loop, whose c...

The aim of the present explorative study was to acoustically examine tarankanje - a special manner of singing typical for the folk music of Istrian-Littoral region. The tarankanje performances of twelve professional male folk singers were digitally recorded and analysed using long-term average spectrum (LTAS) in the PRAAT program. In order to revea...