Models Of Speech Production Research Articles

Characterizing sensorimotor profiles in children with residual speech sound disorder: a pilot study

PurposeChildren with speech errors who have reduced motor skill may be more likely to develop residual errors associated with lifelong challenges. Drawing on models of speech production that highlight the role of somatosensory acuity in updating motor plans, this pilot study explored the relationship between motor skill and speech accuracy, and between somatosensory acuity and motor skill in children. Understanding the connections among sensorimotor measures and speech outcomes may offer insight into how somatosensation and motor skill cooperate during speech production, which could inform treatment decisions for this population. MethodTwenty-five children (ages 9-14) produced syllables in an /ɹ/ stimulability task before and after an ultrasound biofeedback treatment program targeting rhotics. We first tested whether motor skill (as measured by two ultrasound-based metrics of tongue shape complexity) predicted acoustically measured accuracy (the normalized difference between the second and third formant frequencies). We then tested whether somatosensory acuity (as measured by an oral stereognosis task) predicted motor skill, while controlling for auditory acuity. ResultsOne measure of tongue shape complexity was a significant predictor of accuracy, such that higher tongue shape complexity was associated with lower accuracy at pre-treatment but higher accuracy at post-treatment. Based on the same measure, children with better somatosensory acuity produced /ɹ/ tongue shapes that were more complex, but this relationship was only present at post-treatment. ConclusionThe predicted relationships among somatosensory acuity, motor skill, and acoustically measured /ɹ/ production accuracy were observed after treatment, but unexpectedly did not hold before treatment. The surprising finding that greater tongue shape complexity was associated with lower accuracy at pre-treatment highlights the importance of evaluating tongue shape patterns (e.g., using ultrasound) prior to treatment, and has the potential to suggest that children with high tongue shape complexity at pre-treatment may be good candidates for ultrasound-based treatment.

On the Emergence of Phonological Knowledge and on Motor Planning and Motor Programming in a Developmental Model of Speech Production.

A broad sketch for a model of speech production is outlined which describes developmental aspects of its cognitive-linguistic and sensorimotor components. A description of the emergence of phonological knowledge is a central point in our model sketch. It will be shown that the phonological form level emerges during speech acquisition and becomes an important representation at the interface between cognitive-linguistic and sensorimotor processes. Motor planning as well as motor programming are defined as separate processes in our model sketch and it will be shown that both processes revert to the phonological information. Two computational simulation experiments based on quantitative implementations (simulation models) are undertaken to show proof of principle of key ideas of the model sketch: (i) the emergence of phonological information over developmental stages, (ii) the adaptation process for generating new motor programs, and (iii) the importance of various forms of phonological representation in that process. Based on the ideas developed within our sketch of a production model and its quantitative spell-out within the simulation models, motor planning can be defined here as the process of identifying a succession of executable chunks from a currently activated phoneme sequence and of coding them as raw gesture scores. Motor programming can be defined as the process of building up the complete set of motor commands by specifying all gestures in detail (fully specified gesture score including temporal relations). This full specification of gesture scores is achieved in our model by adapting motor information from phonologically similar syllables (adapting approach) or by assembling motor programs from sub-syllabic units (assembling approach).

The neural encoding of productive phonological alternation in speech production: Evidence from Mandarin Tone 3 sandhi

The understanding of alternation is a key goal in phonological research. But little is known about how phonological alternations are implemented in speech production. The current study tested the hypothesis that the production of words that undergo a highly productive alternation, Mandarin Tone 3 sandhi, is supported by a computation mechanism, which predicts that this alternation is subserved by neural activity in a time-window associated with post-lexical phonological and phonetic encoding regardless of word frequency. ERPs were recorded while participants sub-vocally produced high- and low-frequency disyllabic words that do or do not require sandhi. Sandhi words elicited more positive ERPs than non-sandhi words over left anterior channels around 336–520 ms after participants saw the cue instructing them to initiate sub-vocal production, but this effect was not significantly modulated by word frequency. These findings are consistent with predictions of the computation mechanism and have implications for current psycholinguistic models of speech production. (150 words) • ERPs during covert word production can shed light on nature of production process. • T3 sandhi status regulates the ERPs during covert production of Mandarin words. • T3 sandhi effect is observed in time window of phonological and phonetic encoding. • T3 sandhi effect does not interact with word frequency. • Production of productive phonological alternation involves a computation mechanism.

Perceptual Cue Weighting Is Influenced by the Listener's Gender and Subjective Evaluations of the Speaker: The Case of English Stop Voicing.

Speech categories are defined by multiple acoustic dimensions and their boundaries are generally fuzzy and ambiguous in part because listeners often give differential weighting to these cue dimensions during phonetic categorization. This study explored how a listener's perception of a speaker's socio-indexical and personality characteristics influences the listener's perceptual cue weighting. In a matched-guise study, three groups of listeners classified a series of gender-neutral /b/-/p/ continua that vary in VOT and F0 at the onset of the following vowel. Listeners were assigned to one of three prompt conditions (i.e., a visually male talker, a visually female talker, or audio-only) and rated the talker in terms of vocal (and facial, in the visual prompt conditions) gender prototypicality, attractiveness, friendliness, confidence, trustworthiness, and gayness. Male listeners and listeners who saw a male face showed less reliance on VOT compared to listeners in the other conditions. Listeners' visual evaluation of the talker also affected their weighting of VOT and onset F0 cues, although the effects of facial impressions differ depending on the gender of the listener. The results demonstrate that individual differences in perceptual cue weighting are modulated by the listener's gender and his/her subjective evaluation of the talker. These findings lend support for exemplar-based models of speech perception and production where socio-indexical features are encoded as a part of the episodic traces in the listeners' mental lexicon. This study also shed light on the relationship between individual variation in cue weighting and community-level sound change by demonstrating that VOT and onset F0 co-variation in North American English has acquired a certain degree of socio-indexical significance.

Delayed Auditory Feedback Elicits Specific Patterns of Serial Order Errors in a Paced Syllable Sequence Production Task.

Delayed auditory feedback (DAF) interferes with speech output. DAF causes distorted and disfluent productions and errors in the serial order of produced sounds. Although DAF has been studied extensively, the specific patterns of elicited speech errors are somewhat obscured by relatively small speech samples, differences across studies, and uncontrolled variables. The goal of this study was to characterize the types of serial order errors that increase under DAF in a systematic syllable sequence production task, which used a closed set of sounds and controlled for speech rate. Sixteen adult speakers repeatedly produced CVCVCV (C = consonant, V = vowel) sequences, paced to a "visual metronome," while hearing self-generated feedback with delays of 0-250 ms. Listeners transcribed recordings, and speech errors were classified based on the literature surrounding naturally occurring slips of the tongue. A series of mixed-effects models were used to assess the effects of delay for different error types, for error arrival time, and for speaking rate. DAF had a significant effect on the overall error rate for delays of 100 ms or greater. Statistical models revealed significant effects (relative to zero delay) for vowel and syllable repetitions, vowel exchanges, vowel omissions, onset disfluencies, and distortions. Serial order errors were especially dominated by vowel and syllable repetitions. Errors occurred earlier on average within a trial for longer feedback delays. Although longer delays caused slower speech, this effect was mediated by the run number (time in the experiment) and small compared with those in previous studies. DAF drives a specific pattern of serial order errors. The dominant pattern of vowel and syllable repetition errors suggests possible mechanisms whereby DAF drives changes to the activity in speech planning representations, yielding errors. These mechanisms are outlined with reference to the GODIVA (Gradient Order Directions Into Velocities of Articulators) model of speech planning and production. https://doi.org/10.23641/asha.19601785.

How relative frequency and prosodic structure affect the acoustic duration of English derivatives

Morphological segmentability, i.e., the degree to which complex words can be decomposed into their morphological constituents, has been considered an important factor in research on morphological processing and is expected to affect acoustic duration (e.g., Hay, 2001, 2003). One way of operationalizing segmentability is through the relative frequency of a complex word to its base word. However, relative frequency has failed to affect duration for different affix categories in many previous studies. One potential reason is the fact that complex words vary in their prosodic structure, depending on the prosodic integration of the affix (Plag & Ben Hedia, 2018).In a large corpus study with three different corpora and eight affixes each, we investigate how prosodic word structure and relative frequency influence duration, and how these two factors interact. We find that prosodic structure does not significantly interact with relative frequency. Second, we show that relative frequency effects on duration do not emerge consistently across a large number of affixes. Third, not only does prosodic word structure not explain the absence of relative frequency effects, it also often cannot account for durational differences as such. We discuss these findings in light of phonological theory and speech production models.

The Influence of Cognitive, Affective, and Sociocultural Individual Differences on L2 Chinese Speech Performance: A Multiple Indicators Multiple Causes Approach.

Informed by Segalowitz's (Cognitive bases of second language fluency, Routledge, 2010) L2 speech production model and MacIntyre et al.'s (Mod Lang J 82(4):545-562, 1998. https://doi.org/10.1111/j.1540-4781.1998.tb05543.x ) L2 willingness to communicate model, this study sought to understand the influence of cognitive, affective and sociocultural individual differences on advanced learners' L2 Chinese speech performance. A total of 240 advanced L2 Chinese learners in China participated in the study. The participants' perceptions of the impact of cognitive, affective, and sociocultural factors on their L2 Chinese speech performance were measured by an adapted questionnaire. A speaking test, following the Hanyu Shuiping Kouyu Kaoshi (HSKK, an international standardized L2 Chinese speaking proficiency test for non-native speakers), was employed to evaluate the participants' L2 Chinese speech performance. The results of multiple indicators multiple causes (MIMIC) analysis show that (1) cognitive factors such as cognitive fluency, expression practice strategy, and assistance strategy, (2) affective factors such as speaking self-efficacy, speaking anxiety, and speaking motivation, and (3) sociocultural factors such as attitudes toward target language class and attitudes toward target language culture jointly influence advanced L2 Chinese learners' speech performance. Results and implications of the present study are discussed for enhancinglearners' L2 Chinese speech performance.

Perception of English vowel contrasts by Acehnese-Indonesian bilingual learners of English

Previous studies have reported that second language (L2) learners filter non-native L2 sounds through their existing native or first language (L1) sounds when learning L2 sounds. The degree of similarity between L1 and L2 sounds can predict the ease of acquisition of non-native L2 sounds. In the context of English language learning in Indonesia, most learners are likely to speak two languages before they learn English at school; Acehnese is not wituout any exception. As a result, they have larger phonemic inventories to rely on when learning English sounds. This study seeks to investigate how Acehnese-Indonesian bilinguals perceive five sets of vowel contrast in English (i.e., /ɪ/ - /iː/, /æ/ - /ɛ/, /ʌ/ - /ɑː/, /ʊ/ - /uː/, and /ɑː/ - /ɔː/). A special focus is given to their perception of English vowel contrasts that are new, similar, and identical to Acehnese and Indonesian vowels. A group of 31 high school students (N=31) from an Islamic boarding school in Aceh participated in this study. An AX test comprising repetition and minimal pairs of English vowel contrasts in CVC word contexts was randomly presented to the students. The data were analyzed by comparing the means of each vowel pair followed by a general linear model statistical analysis and interpreted based on speech perception and production models. The findings indicate that the Acehnese-Indonesian bilinguals discriminate the vowel pairs /æ/ - /ɜː/ and /ɑː/ - /ɔː/ better than the /ɪ/ - /iː/, /ʌ/ - /ɑː/ and /ɑː/ - /ɔː/ pairs. The vowel pairs in which both vowels are novel to Acehnese and Indonesian were moderately discriminated compared to the pairs in which one vowel is similar in both Acehnese and Indonesian. It can be said that students perceived pairs with one vowel similar to Acehnese-Indonesian better than pairs with both vowels or one vowel novel in Acehnese and Indonesian. The findings reported in the paper are expected to inform pedagogic practices, particularly in the development of materials for the teaching of English pronunciation. English teachers in Aceh may incorporate Acehnese words that have similar sounds to English while emphasizing the novel sounds which are absent from the Acehnese vowel system.

МОДЕЛЬ ПОРОЖДЕНИЯ РЕЧИ ШКОЛЫ Л.С. ВЫГОТСКОГО: ОСНОВЫ И ВЕРИФИКАЦИЯ

In this article, the development and verification of the A.A. Leontiev–T.V. Akhutina model of speech production based on L.S. Vygotsky’s ideas is discussed. The conceptual and experimental background of L.S. Vygotsky’s theory of speech thinking is analyzed. The ideas of this theory opened the way to a prudent approach to understanding the controversy in psycholinguistics of the 1960s–1970s (works by A.A. Leontiev), A.R. Luria’s findings in aphasiology, and the results of T.V. Akhutina’s research in the field of agrammatism. This determined, in many ways, the specifics of the general framework of the model. The major psycholinguistic concepts elaborated in the 1970s are considered, i.e., those introduced by Ch. Fillmore, C. Lee and S. Thompson, J. Bruner and E. Bates, all shedding light on the syntactic operations of the model. The verification of these ideas in the studies describing syntax in children’s speech, “pidgin” and creole languages, etc. is carried out. The recent observations of the problem being studied at the intersection of linguistics, psychology, and neurosciences are summarized. Particular attention is paid to R. Jackendoff and E. Wittenberg’s review of “linear” languages, as well as to the experimental studies by R. Tomlin, A. Myachikov, A. Konopka that fit with the A.A. Leontiev–T.V. Akhutina model of speech production. The conclusion is made about the high explanatory potential of L.S. Vygotsky’s ideas and the model based on them.

Out of sight, out of mind: The influence of communicative load and phonological neighborhood density on phonetic variation in real listener-directed speech.

Some models of speech production propose that speech variation reflects an adaptive trade-off between the needs of the listener and constraints on the speaker. The current study considers communicative load as both a situational and lexical variable that influences phonetic variation in speech to real interlocutors. The current study investigates whether the presence or absence of a target word in the sight of a real listener influences speakers' patterns of variation during a communicative task. To test how lexical difficulty also modulates intelligibility, target words varied in phonological neighborhood density (ND), a measure of lexical difficulty. Acoustic analyses reveal that speakers produced longer vowels in words that were not visually present for the listener to see, compared to when the listener could see those words. This suggests that speakers assess in real time the presence or absence of supportive visual information in assessing listener comprehension difficulty. Furthermore, the presence or absence of the word interacted with ND to predict both vowel duration and hyperarticulation patterns. These findings indicate that lexical measures of a word's difficulty and speakers' online assessment of lexical intelligibility (based on a word's visual presence or not) interactively influence phonetic modifications during communication with a real listener.

Формирование устного дискурса у детей 4—5 лет: функционально-динамический анализ влияния жанра и регистра на грамматический частотный профиль

Despite a long history of studies in the discourse, we still lack systematic investigations in the discourse acquisition at the preschool and pre-primary school age. Namely, the influence of variables, such as discourse genre, communication register, socio-pragmatic skills, on different linguistic parameters almost has not been investigated yet. Hence, the given study aimed at the corpus-based analysis of different child discourse genres with the main focus on the Part-of-Speech (PoS) Profile. First, each child told three personal narratives, then generated a fictional story according to the picture sequence, and had an informal chat during five individual sessions. Then, five subcorpora, including different genres (story vs. conversation) and different registers (speech addressed to adult vs. peer), were developed to analyze the distribution of parts of speech. Finally, individual and within-group PoS Profiles were built. Statistical analysis revealed some significant differences in the PoS Profile between the genres and communication registers. Namely, the genre had a significant impact on the frequency of nouns, verbs, adverbs, particles, conjunctions, and prepositions, while the communication register impacted significantly on the frequency of nouns, pronouns, adverbs, particles, and conjunctions. In the paper, the dynamic model of speech production is further discussed. Results of the study inspire discussion on the role of the PoS Profile as a manifestation of the concurrent relations between the child’s language competence and cognitive resources.

Significant pathological voice discrimination by computing posterior distribution of balanced accuracy

The ability to speak lucidly plays a key role in social relations. Consequently, the role of the larynx is quite important, and timely diagnosis of laryngeal diseases has proved to be crucial. In this study, a simple computational model for inverse of speech production model is employed to extract the glottal waveform using speech signal. This waveform has useful information about vocal folds performance in terms of providing evidence for distinguishing pathological disorders. Furthermore, obtaining the significance of classification results is important, because it leads to reliable inferences. This study utilizes the sustained vowel sound /a/ and a well-referenced database, namely MEEI. In this work, after extraction of six discriminating features by using appropriate signal modeling and processing methods and upon change of the feature space by using Kernel Principal Component Analysis (KPCA), a classifier consisting of Naïve Bayes and Fisher Linear Discriminant, is exerted on the feature sets. Regarding voice pathology detection, the proposed approach achieved a significant classification balanced accuracy 93.6%±0.03 with p-value <0.01 for normal and abnormal classification using the Beta distribution model for the posterior distribution of the average of the cross-validation results. The proposed features are also compared with some conventional features in this field. The results show significant improved performance for the proposed features in discriminating different types of pathological voices.

The relation of velopharyngeal coupling area to the identification of stop versus nasal consonants in North American English based on speech generated by acoustically driven vocal tract modulations

The purpose of this study was to determine the threshold of velopharyngeal coupling area at which listeners switch from identifying a consonant as a stop to a nasal in North American English, based on V1CV2 stimuli generated with a speech production model that encodes phonetic segments as relative acoustic targets. Each V1CV2 was synthesized with a set of velopharyngeal coupling functions whose area ranged from 0 to 0.1 cm2. Results show that consonants were identified by listeners as a stop when the coupling area was less than 0.035-0.057 cm2, depending on place of articulation and final vowel. The smallest coupling area (0.035 cm2) at which the stop-to-nasal switch occurred was found for an alveolar consonant in the /ɑCi/ context, whereas the largest (0.057 cm2) was for a bilabial in /ɑCɑ/. For each stimulus, the balance of oral versus nasal acoustic energy was characterized by the peak nasalance during the consonant. Stimuli with peak nasalance below 40% were mostly identified by listeners as stops, whereas those above 40% were identified as nasals. This study was intended to be a precursor to further investigations using the same model but scaled to represent the developing speech production system of male and female talkers.

The duration of word-final /s/ differs across morphological categories in English: evidence from pseudowords.

Previous research suggests that different types of word-final /s/ and /z/ (e.g. non-morphemic vs. plural or clitic morpheme) in English show realisational differences in duration. However, there is disagreement on the nature of these differences, as experimental studies have provided evidence for durational differences of the opposite direction as results from corpus studies (i.e. non-morphemic>plural>clitic /s/). The experimental study reported here focuses on four types of word-final /s/ in English, i.e. non-morphemic, plural, and is- and has-clitic /s/. We conducted a pseudoword production study with native speakers of Southern British English. The results show that non-morphemic /s/ is significantly longer than plural /s/, which in turn is longer than clitic /s/, while there is no durational difference between the two clitics. This aligns with previous corpus rather than experimental studies. Thus, the morphological category of a word-final /s/ appears to be a robust predictor for its phonetic realisation influencing speech production in such a way that systematic subphonemic differences arise. This finding calls for revisions of current models of speech production in which morphology plays no role in later stages of production.

Pitch accent disorder observed in a Japanese patient due to traumatic head injury

ABSTRACT Background Suprasegmental features are often affected by brain damage, and are typically accompanied by motor speech disorders such as dysarthria, apraxia of speech (AOS), and foreign accent syndrome (FAS). However, there is no report of a selective pitch accent disorder without motor speech impairments or FAS. Aims We describe a female Japanese patient who developed a selective pitch accent disorder without motor speech impairments or FAS, with accompanying mild Wernicke’s aphasia due to head injury, in order to investigate the mechanism of pitch accent disorder based on speech production models. Methods & Procedures We analysed her spontaneous speech, confrontation naming, repetition, reading aloud, and the error patterns in her pitch accents in two sessions. Additionally, we gave her a reading aloud test of homonyms with different accent patterns and an auditory discrimination task of pitch accents. Outcomes & Results The patient exhibited a significant recovery from aphasia, and paraphasia was observed to almost disappear, although the pitch accent errors remained in all speech modalities. While she was able to auditorily distinguish between correct and incorrect pitch accents, she was unaware of her own abnormally accented speech. The results of the analysis of the error patterns in her pitch accents revealed that they, even if incorrect for the selected words, did not actually deviate from the correct patterns in the Osaka dialect, her native dialect. Conclusions Our observations suggest that the process of pitch accent production may be independent of that of phoneme production. Moreover, it may be that her accent abnormalities were not produced randomly during the process of articulation, but instead may have been due to a problem in selecting pitch accent patterns from the repertoire of Osaka accent patterns allocated to the corresponding words. This suggests that the patient’s pitch accent disorder was a problem in a higher level of linguistic processing, i.e., lexical pitch accent encoding, rather than the motor control of speech production.

Robustness and sensitivity metrics-based tuning of the augmented Kalman filter for single-channel speech enhancement

The inaccurate estimates of the speech and noise linear prediction coefficients (LPCs) introduce bias in augmented Kalman filter (AKF) gain, which impacts the quality and intelligibility of enhanced speech. Although current tuning methods offset the bias in AKF gain, particularly in colored noise conditions, they do not adequately address nonstationary noise conditions. This paper introduces a new tuning algorithm of the AKF gain for speech enhancement in real-life noise conditions. Due to this purpose, a speech presence probability (SPP) method first estimates the noise power spectral density (PSD) from each noisy speech frame to compute the noise LPC parameters. A whitening filter is constructed with the noise LPCs to pre-whiten each noisy speech frame prior to computing the speech LPC parameters. The AKF is then constructed with the estimated speech and noise LPC parameters. To achieve better noise reduction, the robustness metric is employed to dynamically offset the bias in AKF gain during speech absence of the noisy speech to that of the sensitivity metric during speech presence. The speech activity is obtained through adopting the speech and noise production model parameters. It is shown that the reduced-biased AKF gain achieved by the proposed tuning algorithm addresses speech enhancement in real-life noise conditions. Objective and subjective scores on the NOIZEUS corpus demonstrate that the proposed method produces enhanced speech with higher quality and intelligibility than the competing methods in real-life noise conditions for a wide range of signal-to-noise ratio (SNR) levels.

Articulatory and acoustic analysis of voice acting

Voice actors are an under-utilized population for linguistic study. Looking at voice actors' novel productions of typical speech sounds can tell us more about over which muscles we have volitional control. This can help test models of speech production, inform speech therapy, and create better pedagogy for teaching vocal performance. While there have been several studies analyzing the acoustics produced by voice actors (Teshigawara, 2001; 2003; 2004; 2009; 2011; Teshigawara et al., 2007; Uchida, 2007) there are fewer studies that have investigated the articulation of voice acting (Teshigawara &amp; Murano, 2004). This study used acoustic analysis, 3D ultrasound, and electroglottography to investigate the speech produced by six professional and amateur voice actors (2 female, 4 male). Measures of average fundamental frequency (F0) and F0 range were collected from each actor speaking in their regular adult voice and a simulated child voice. 3D ultrasound was used to look at tongue placement and horizontal kymograms of the ultrasound images were also used to infer hyoid bone movement. Additionally, acoustic estimates of vocal tract length were compared to electroglottography measurements of laryngeal raising for two of the subjects. This presentation will describe the within-subject variation across each actor’s adult and simulated child voices.

L2 pronunciation research and teaching

Abstract Research on L2 pronunciation needs to occur in a much wider range of languages in order to ensure that findings are not just applicable to dominant languages. This paper argues that research on a wider variety of languages will be valuable for understanding different contexts of learning/instruction, different types and combinations of phonological features, and different instructional approaches. The field especially needs two parallel and interrelated threads of research and teaching-oriented publications. The first is descriptions of and materials for teaching pronunciation in particular languages. The second is research that is built around questions of practical interest and theoretically-motivated models of L2 speech perception and production.

Automation of Function Assignment in the Models of Speech Production and Second Language Acquisition

This article explores the concept of function assignment in first language (L1) and second language (L2) speech production, compares automation of function assignment in L1 and L2 speech production, pursues factors hampering automation of function assignment in L2 speech production, and discusses how to improve automation of function assignment in L2 speech production. Grammatical encoding, of which function assignment is one of the subordinate processes, is also one of the processes in L2 speech production. While function assignment in L1 speech production is mostly automatic, it demands much attentional resources and is executed under conscious supervision in L2 speech production. L2 learners’ incomplete knowledge of the target language and their limited working memory resources hamper automation of function assignment in L2 speech production. Furthermore, as per generative models of learning, to improve automation of function assignment, teachers can either adopt strategies or improve instructional designs targeting this subprocess. Together, this conceptual paper gives a comprehensive overview of automation of function assignment, explores its impact on second language acquisition (SLA), and reveals that it is feasible to facilitate automation of function assignment in L2 speech production by improving instructional designs, especially the presentation methods of sentence elements.

Morpho-Phonetic Effects in Speech Production: Modeling the Acoustic Duration of English Derived Words With Linear Discriminative Learning.

Recent evidence for the influence of morphological structure on the phonetic output goes unexplained by established models of speech production and by theories of the morphology-phonology interaction. Linear discriminative learning (LDL) is a recent computational approach in which such effects can be expected. We predict the acoustic duration of 4,530 English derivative tokens with the morphological functions DIS, NESS, LESS, ATION, and IZE in natural speech data by using predictors derived from a linear discriminative learning network. We find that the network is accurate in learning speech production and comprehension, and that the measures derived from it are successful in predicting duration. For example, words are lengthened when the semantic support of the word's predicted articulatory path is stronger. Importantly, differences between morphological categories emerge naturally from the network, even when no morphological information is provided. The results imply that morphological effects on duration can be explained without postulating theoretical units like the morpheme, and they provide further evidence that LDL is a promising alternative for modeling speech production.