Monthly Archives: April 2010
Modeling the contribution of phonotactic cues to the problem of word segmentation*
How do infants find the words in the speech stream? Computational models help us understand this feat by revealing the advantages and disadvantages of different strategies that infants might use. Here, we outline a computational model of word segmentation that aims both to incorporate cues proposed by language acquisition researchers and to establish the contributions different cues can make to word segmentation. We present experimental results from modified versions of Venkataraman’s (2001) segmentation model that examine the utility of: (1) language-universal phonotactic cues; (2) language-specific phonotactic cues which must be learned while segmenting utterances; and (3) their combination. We show that the language-specific cue improves segmentation performance overall, but the language-universal phonotactic cue does not, and that their combination results in the most improvement. Not only does this suggest that language-specific constraints can be learned simultaneously with speech segmentation, but it is also consistent with experimental research that shows that there are multiple phonotactic cues helpful to segmentation (e.g. Mattys, Jusczyk, Luce & Morgan, 1999; Mattys & Jusczyk, 2001). This result also compares favorably to other segmentation models (e.g. Brent, 1999; Fleck, 2008; Goldwater, 2007; Johnson & Goldwater, 2009; Venkataraman, 2001) and has implications for how infants learn to segment.
from the Journal of Child Language
Segmenting words from natural speech: subsegmental variation in segmental cues*
Most computational models of word segmentation are trained and tested on transcripts of speech, rather than the speech itself, and assume that speech is converted into a sequence of symbols prior to word segmentation. We present a way of representing speech corpora that avoids this assumption, and preserves acoustic variation present in speech. We use this new representation to re-evaluate a key computational model of word segmentation. One finding is that high levels of phonetic variability degrade the model’s performance. While robustness to phonetic variability may be intrinsically valuable, this finding needs to be complemented by parallel studies of the actual abilities of children to segment phonetically variable speech.
from the Journal of Child Language
Words in puddles of sound: modelling psycholinguistic effects in speech segmentation*
There are numerous models of how speech segmentation may proceed in infants acquiring their first language. We present a framework for considering the relative merits and limitations of these various approaches. We then present a model of speech segmentation that aims to reveal important sources of information for speech segmentation, and to capture psycholinguistic constraints on children’s language perception. The model constructs a lexicon based on information about utterance boundaries and deduces phonotactic constraints from the discovered lexicon. Compared to other models of speech segmentation, our model performs well in terms of accuracy, computational tractability and the number of components of the model. Finally, our model also reflects the psycholinguistic effects of language learning, in terms of the early advantage for segmentation provided by the child’s name, and by revealing the overlap in usefulness of information for segmentation and for grammatical categorization of the language.
from the Journal of Child Language
An image is worth a thousand words: why nouns tend to dominate verbs in early word learning
Nouns are generally easier to learn than verbs (e.g. Bornstein, 2005; Bornstein et al., 2004; Gentner, 1982; Maguire, Hirsh-Pasek & Golinkoff, 2006). Yet, verbs appear in children’s earliest vocabularies, creating a seeming paradox. This paper examines one hypothesis about the difference between noun and verb acquisition. Perhaps the advantage nouns have is not a function of grammatical form class but rather is related to a word’s imageability. Here, word imageability ratings and form class (nouns and verbs) were correlated with age of acquisition according to the MacArthur-Bates Communicative Development Inventory (CDI; Fenson et al., 1994). CDI age of acquisition was negatively correlated with words’ imageability ratings. Further, a word’s imageability contributes to the variance of the word’s age of acquisition above and beyond form class, suggesting that at the beginning of word learning, imageability might be a driving factor.
Direct Instruction in Written Expression: The Effects on English Speakers and English Language Learners with Disabilities
Students with disabilities often struggle with writing tasks. In order to improve the written expression performance of high school students with deficits in written expression, a Direct Instruction writing program was implemented. The participants were six high school students in programs for individuals with learning disabilities. Three of the six students were served in programs for students who are English Language Learners. Using a multiple-probe across-participants design, the effect of the writing program was examined. The intervention was implemented over a 5-week period with maintenance checks conducted 2 and 4 weeks after the termination of instruction. Results were variable, but there appeared to be a positive trend in student writing performance as measured by correct word sequence, length of text, and the TOWL-3. Implications for practice and future directions are also provided.
The Effects of a Fluency Intervention Program on the Fluency and Comprehension Outcomes of Middle-School Students with Severe Reading Deficits
Despite advances in the science of teaching reading, there still exists a small percentage of students who fail to make the expected progress in reading-related skills, notwithstanding attempts at intervention. Even if these struggling readers learn to decode adequately, fluency remains a problem for many, and little is known about the effectiveness of fluency interventions for older students with severe reading deficits. This study used a randomized experimental design to test the efficacy of a fluency intervention program on the word-identification and reading-comprehension outcomes of 60 middle-school students with severe reading delays. Results showed that students in the experimental group made more progress on standardized tests of reading fluency than students in the control group. No gains were seen in reading comprehension.
Measuring Attitudes Toward Stuttering: English-to-French Translations in Canada and Cameroon
Purpose. A field test of a survey instrument under development, the Public Opinion Survey of Human Attributes, Experimental Edition (POSHA-E), designed to investigate language-, culture-, and nation-specific public opinions about stuttering is reported. This investigatiocompared English and French versions of the POSHA-E in widely disparate cultures to explore culturversus language influences. Method. 120 experimental respondents rated POSHA-E items on 1- 9 equappearing interval scales: 30 in English and 30 in French in both Canada and Cameroon. Comparisons were made with 30 matched, monolingual English, American respondents in English only. Results. Between-country differences for stuttering between experimental groups were much larger than between-language differences. Conclusions. The POSHA-E can be translated to another language, i.e. French, without significant change in item meaning and interpretation in two divergent cultures, advancing the development and validity of an instrument that can be used in different language and cultural settings worldwide.
from the Journal of Communication Disorders
An fMRI investigation of cognitive stages in reasoning by analogy
We compared reasoning about four-term analogy problems in the format (A:B :: C: D) to semantic and perceptual control conditions that required matching without analogical mapping. We investigated distinct phases of the problem solving process divided into: encoding, mapping/inference, and response. Using fMRI, we assessed the brain activation relevant to each of these phases with an emphasis on achieving a better understanding of analogical reasoning relative to these other matching conditions. We predicted that the analogical condition would involve the most cognitive effort in the encoding and mapping/inference phases, while the control conditions were expected to engage greater prefrontal cortex (PFC) activation at the response period. Results showed observed greater activation for the analogical condition relative to the control conditions at the encoding phase in several predominantly left lateralized and medial areas of the PFC. Similar results were observed for the mapping/inference phase, though this difference was limited to the left PFC and rostral PFC. The response phase showed fastest and most accurate responses in the analogy condition relative to the control conditions. This was accompanied by greater processing within the left lateral and the medial PFC for the control conditions relative to the analogy condition, consistent with most of the cognitive processing of the analogy condition having occurred in the prior task phases. Overall we demonstrate that the left ventral and dorsal lateral, medial, and rostral PFC are important in both the encoding of relational information, mapping and inference processes, and verification of semantic and perceptual responses in four term analogical reasoning.
from Brain Research
Effects of Practice Variability on Learning of Relaxed Phonation in Vocally Hyperfunctional Speakers
The present study investigated the effects of practice variability on the learning of relaxed phonation using a motor learning perspective. Twenty-one individuals with hyperfunctional voice problems were evenly and randomly assigned to three groups of practice conditions: constant, blocked, and random practice conditions. During training, participants in the constant practice condition were asked to read aloud sentence stimuli with four Chinese characters. Participants in the blocked practice condition were asked to read aloud sentence stimuli with increasing sentence length, starting from sets of two characters to five characters. Participants in the random practice condition were asked to practice reading sentence stimuli of variable length from two to five characters presented in a random fashion. Surface electromyographic feedback (sEMG) from the thyrohyoid muscle site was given to each participant after reading every two sentence stimuli. Results demonstrated that for all the participants, voice motor learning was evidenced by the decreased sEMG levels in delayed retention test. Generalization to untrained passage was shown as well. However, results did not reveal any difference in the learning among the three practice conditions. The findings from the present study did not support the hypothesis of contextual interference, which states that practice using variable items presented in a random mode is more beneficial to learning than practice using constant items.
from the Journal of Voice
Formation of the Actor’s/Speaker’s Formant: A Study Applying Spectrum Analysis and Computer Modeling
A 1D model-based optimization suggested that this kind of a formant cluster could be best established by simultaneously narrowing the epilaryngeal tube, widening the pharynx and narrowing the front of the oral cavity.
from the Journal of Voice
Listener Perception of the Effect of Abdominal Kinematic Directives on Respiratory Behavior in Female Classical Singing
Breath management training in classical singing is becoming increasingly physiologically focused, despite evidence that directives focusing on chest-wall kinematic (ribcage and abdominal) behavior effect minimal change in acoustical measures of singing. A direct and proportionate relationship between breathing behavior and vocal quality is important in singing training because singing teachers rely primarily on changes in sound quality to assess the efficacy of breath management modification. Pedagogical opinion is also strongly divided over whether the strategy of retarding the reduction in abdominal dimension during singing has a negative effect on vocal quality. This study investigated whether changes in abdominal kinematic strategy were perceptible and whether listeners preferred a particular strategy. Fourteen experienced singing teachers and vocal coaches assessed audio samples of five female classical singers whose respiratory kinematic patterns during singing had been recorded habitually and under two simple, dichotomous directives: Gradually drawing the abdomen inward and gradually expanding the abdomen, during each phrase. Listeners rated the singers on standard of singing and of breath management. Ratings analysis took into consideration changes in kinematic behavior under each directive determined from the respiratory recordings. Listener ratings for two singers were unaffected by directive. For three singers, ratings were lower when the directive opposed habitual kinematic behavior. The results did not support the pedagogical assumption of a direct and proportional link between respiratory behavior and standard of singing or that the abdomen-outward strategy was deleterious to vocal quality. The findings demonstrate the importance of considering habitual breathing behavior in both research and pedagogical contexts.
from the Journal of Voice
Voice Quality in Relation to Voice Complaints and Vocal Fold Condition During the Screening of Female Student Teachers
The present study showed that perceptual assessment of the voice and voice complaints is not sufficient to check if the future professional is at risk. Therefore, preventive measures are needed to detect students at risk early in their education and this depends on broader assessment: on the one hand, assessing voice quality and voice complaints and on the other hand, examination of the vocal folds of all starting students.
from the Journal of Voice
Five-Year Longitudinal Treatment Outcomes of the ISTAR Comprehensive Stuttering Program
Replicated evidence of satisfactory 1- and 2-year post-treatment outcomes has been reported for the Comprehensive Stuttering Program (CSP). However, little is known about longer-term outcomes of the CSP. Yearly follow-up measures were obtained from 18 participants for 5 consecutive years. At 5-years follow-up, participants were maintaining clinically and statistically significant reductions in stuttering and increases in rates of speech relative to pre-treatment measures. Standardized effect sizes were large. There were no significant differences among the immediate post-treatment and five follow-up measures, indicating that speech gains achieved by the end of the treatment program were stable over the 5-year follow-up period. Insufficient return rates for self-report data for the third to fifth follow-up measurement occasions prohibited analyzing these data. However, non-significant differences among the immediate post-treatment and two follow-up measures indicated that improvements achieved by the end of treatment in speech related confidence, and perceptions of struggle, avoidance, and expectancy to stutter were stable over the 2-year follow-up period. Significant differences among the speech related communication attitudes scores indicated that improvements in attitudes made at the end of the treatment program were less stable. Taken together, these results provide further and longer-term evidence of the effectiveness of the CSP.
Redesigning therapy for agrammatism: Initial findings from the ongoing evaluation of a conversation-based intervention study
Therapy for agrammatism, once only targeted at surface grammar, has begun to demonstrate the benefits of rigorous theoretically motivated therapy aimed at underlying syntax. Whilst there is evidence that grammatical ability in the clinical setting can be improved by such therapies, it has proved hard to detect carryover to everyday conversation in the home. Recent research using a qualitative methodology called Conversation Analysis (CA) has shown that utterances produced by agrammatic speakers in peer conversation differ significantly from those elicited during assessment and therapy tasks. This is because tasks target decontextualized language, isolated from an interactional context of real-life talk about needs, opinions, and experiences. This paper explores the idea that, given this finding, it may be more appropriate, and effective, to provide therapy for agrammatism by targeting the grammar of conversation directly. It outlines a new therapeutic approach based on CA concepts, describes the early development of a valid quantitative measure of change in conversation and presents some qualitative data from one of the first dyads involved in an ongoing therapy study.
from the Journal of Neurolinguistics
Longitudinal cerebral blood flow changes during speech in hereditary ataxia
The hereditary ataxias constitute a group of degenerative diseases that progress over years or decades. With principal pathology involving the cerebellum, dysarthria is an early feature of many of the ataxias. Positron emission tomography was used to study regional cerebral blood flow changes during speech production over a 21 month period in a group of seven right-handed subjects with hereditary ataxia (6 females and 1 male, 3 SCA1 and 4 SCA5, aged 38.3 ± 18.9 years). The decline in blood flow was greatest in cerebellar regions. In contrast, blood flow actually increased during speech production in the classic speech area (Broca’s area) but not in its right-hemisphere homologue at the second evaluation. This increase in cortical flow may have been compensatory for cerebellar degeneration as speech intelligibility did not decline significantly during this period. Compensation was not complete, though, as syllable timing shifted in the direction of equal syllable duration, one of the characteristics of ataxic dysarthria. These results are consistent with previous functional imaging studies of ataxia demonstrating a pattern of brain activity that reflects both loss of function and relative compensation when clinical signs and symptoms are still mild. The combination of disease-relevant tasks, behavioral measurement, and functional imaging may provide insight into the early changes associated with neurodegenerative disease.
from Brain and Language
