Grapheme-phoneme Conversion Research Articles

Semantic typicality effects in acquired dyslexia: Evidence for semantic impairment in deep dyslexia

Background: Acquired deep dyslexia is characterised by impairment in grapheme-phoneme conversion and production of semantic errors in oral reading. Several theories have attempted to explain the production of semantic errors in deep dyslexia, some proposing that they arise from impairments in both grapheme-phoneme and lexical-semantic processing, and others proposing that such errors stem from a deficit in phonological production. Whereas both views have gained some acceptance, the limited evidence available does not clearly eliminate the possibility that semantic errors arise from a lexical-semantic input-processing deficit. Aims: To investigate semantic processing in deep dyslexia this study examined the typicality effect in deep dyslexic individuals, phonological dyslexic individuals, and controls using an online category verification paradigm. This task requires explicit semantic access without speech production, focusing observation on semantic processing from written or spoken input. Methods & Procedures: To examine the locus of semantic impairment, the task was administered in visual and auditory modalities with reaction time as the primary dependent measure. Nine controls, six phonological dyslexic participants, and five deep dyslexic participants completed the study. Outcomes & Results: Controls and phonological dyslexic participants demonstrated a typicality effect in both modalities, while deep dyslexic participants did not demonstrate a typicality effect in either modality. Conclusions: These findings suggest that deep dyslexia is associated with a semantic processing deficit. Although this does not rule out the possibility of concomitant deficits in other modules of lexical-semantic processing, this finding suggests a direction for treatment of deep dyslexia focused on semantic processing.

Different letter-processing strategies in diagnostic subgroups of developmental dyslexia also occur in a transparent orthography: Reply to a commentary by Spinelli et al.

The article was motivated by a commentary of Spinelli et al. (2010), who commented on our experimental study with dyslexic children (Lachmann & van Leeuwen, 2008). They questioned the unusually large reversed lexicality effect we reported for three of our dyslexic children for which word reading times were considerably longer than nonword reading times. We argued that, in principle, in a transparent orthography, such as German, children exist who have significant problems in word reading, but for whom nonword reading is normal. The extreme reversed lexically effect, however, may not be representative for the dyslexic population. Since we do not want to give the impression that our results were based on these three participants, we reran analyses on reaction times presented in Lachmann and van Leeuwen, this time excluding the data from the three individuals. Results were replicated. The constructive criticism has helped put both the diagnostics and our experimental results on even firmer ground. Both yield a consistent interpretation, in which two subgroups of dyslexics can be distinguished: one with generic activation problems; the other with a specific problem in phoneme-grapheme conversion.

A novel method for examining response to spelling treatment

Background: According to a dual-route model of written language processing, spelling of irregular words provides an index of the status of lexical spelling procedures, whereas nonword spelling provides information about non-lexical processing that relies on phoneme-grapheme conversion. Because regular words can be spelled using either route, accuracy for such words may reflect the combined function of the two routes, and may be mathematically predicted on the basis of spelling accuracy for irregular words and nonwords. Aims: The purpose of the present study was to examine the application of a dual-route prediction equation, and a related multiple regression model, to evaluate pre-post treatment spelling performance of individuals with acquired alexia/agraphia. Methods & Procedures: Eight individuals with language impairment due to left hemisphere damage received behavioral treatment to improve their written spelling. Their spelling performance was examined before and after treatment on untrained word lists with regular and irregular spellings, and pronounceable nonwords, and concurrence between predicted and observed spelling of regular words was evaluated. Outcomes & Results: The group demonstrated significantly improved spelling performance after treatment, and the prediction equation and multiple regression model both accurately predicted regular word performance on the basis of irregular word and nonword scores. In addition, the multiple regression model provided potentially useful information about the relative contribution of the lexical and non-lexical routes to spelling performace. Conclusions: The prediction equation and related multiple regression model used in this study can offer novel insight into the cognitive processes available to individuals with agraphia as well as provide a quantitative means to characterize response to treatment.

Differential activation of frontal and parietal regions during visual word recognition: An optical topography study

The present study examined cortical oxygenation changes during lexical decision on words and pseudowords using functional Near-Infrared Spectroscopy (fNIRS). Focal hyperoxygenation as an indicator of functional activation was compared over three target areas over the left hemisphere. A 52-channel Hitachi ETG-4000 was used covering the superior frontal gyrus (SFG), the left inferior parietal gyrus (IPG) and the left inferior frontal gyrus (IFG). To allow for anatomical inference a recently developed probabilistic mapping method was used to determine the most likely anatomic locations of the changes in cortical activation [Tsuzuki, D., Jurcak, V., Singh, A.K., Okamoto, M., Watanabe, E., Dan, I., 2007. Virtual spatial registration of stand-alone fNIRS data to MNI space. NeuroImage 43 (4), 1506–1518.Subjects made lexical decisions on 50 low and 50 high frequency words and 100 pseudowords. With respect to the lexicality effect, words elicited a larger focal hyperoxygenation in comparison to pseudowords in two regions identified as the SFG and left IPG. The SFG activation difference was interpreted to reflect decision-related mechanisms according to the Multiple Read-Out Model [Grainger, J., Jacobs, A.M., 1996. Orthographic processing in visual word recognition: A multiple read-out model. Psychological Review 103, 518–565]. The greater oxygenation response to words in the left IPG suggests that this region connects orthographic, phonological and semantic representations. A decrease of deoxygenated hemoglobin was observed to low frequency in comparison to high frequency words in a region identified as IFG. This region’s sensitivity to word frequency suggests its involvement in grapheme–phoneme conversion, or its role during the selection of pre-activated semantic candidates.

Reading and spelling acquisition in European Portuguese: a preliminary study

We investigated the initial development of reading and spelling in European Portuguese. First-graders, tested in February and June, had to read and spell words and pseudowords. In February there were regularity and graphemic complexity effects, indicating that these children relied on grapheme–phoneme conversion. The lexicality effect found in spelling, in June, suggest that by the end of first grade these children had begun to construct an orthographic lexicon. However, lexical addressing is not inconsistent with phonological mediation as regularization errors increased between the sessions. Additionally, the previously reported similarity in global performance of Portuguese and French beginning readers may conceal processing differences that are related to specific characteristics of the corresponding orthographic codes.

Do deep dyslexia, dysphasia and dysgraphia share a common phonological impairment?

This study directly compared four patients who, to varying degrees, showed the characteristics of deep dyslexia, dysphasia and/or dysgraphia—i.e., they made semantic errors in oral reading, repetition and/or spelling to dictation. The “primary systems” hypothesis proposes that these different conditions result from severe impairment to a common phonological system, rather than damage to task-specific mechanisms (i.e. grapheme-phoneme conversion). By this view, deep dyslexic/dysphasic patients should show overlapping deficits but previous studies have not directly compared them. All four patients in the current study showed poor phonological production across different tasks, including repetition, reading aloud and spoken picture naming, in line with the primary systems hypothesis. They also showed severe deficits in tasks that required the manipulation of phonology, such as phoneme addition and deletion. Some of the characteristics of the deep syndromes – namely lexicality and imageability effects – were typically observed in all of the tasks, regardless of whether semantic errors occurred or not, suggesting that the patients’ phonological deficits impacted on repetition, reading aloud and spelling to dictation in similar ways. Differences between the syndromes were accounted for by variation in other primary systems—particularly auditory processing. Deep dysphasic symptoms occurred when the impact of phonological input on spoken output was disrupted or reduced, either as a result of auditory/phonological impairment, or for patients with good phonological input analysis, when repetition was delayed. ‘Deep’ disorders of reading aloud, repetition and spelling can therefore be explained in terms of damage to interacting primary systems such as phonology, semantics and vision, with phonology playing a critical role.

Assignment of phonemes to the graphemes producing them

The assignment of phonemes to graphemes producing them in a lexicon having words (grapheme sequences) and their associated phonetic transcription (phoneme sequences) for the preparation of patterns for training neural networks for the purpose of grapheme-phoneme conversion is carried out with the aid of a variant of dynamic programming which is known as dynamic time warping (DTW).

Can One Suppress Subliminal Words?

Subliminal words cause behavioral priming, yet the depth of their processing remains debated. Using transcranial magnetic stimulation (TMS), Nakamura et al. demonstrate in this issue of Neuron that this subliminal priming effect can be selectively disrupted. Distinct TMS sites disrupt priming in lexical decision and pronunciation tasks, suggesting that task set influences subliminal processing.

Interindividual uniformity and variety of the “Writing center”: A functional MRI study

Our aim is to investigate the neural substrates for writing using fMRI (twenty right-handed subjects). We assumed that common areas involved in both writing with right and left hands are crucial to the central process of writing. We employed Japanese phonograms ( Kana), in which phoneme–grapheme conversion would be extremely simple. Brain activation was examined under three conditions: (1) written naming with the right hand (WR), (2) written naming with the left hand (WL), and (3) naming silently (NA). While the comparison of WR to NA (WR > NA) exhibited activation only in the left frontoparietal area, the WL > NA comparison exhibited broader activation than the WR > NA comparison, i.e., the left frontoparietal area except the motor and sensory areas and the right frontoparietal area. A conjunction analysis in SPM2 revealed common areas of activation across the WR > NA and WL > NA comparisons, which are assumed to be crucial to writing. In the group analysis, three areas were found to be activated: the posterior end of the left superior frontal gyrus, which is superior and posterior to Exner's center; the anterior part of the left superior parietal lobule; and the lower part of the anterior limb of the left supramarginal gyrus. In the single-subject analysis, whereas the first two of the above three areas were found to be crucial for writing in all individuals, an interindividual inconsistency of involvement with writing was observed in three areas: the lower part of the anterior limb of the left supramarginal gyrus (60% involved); the right frontal region (47%); and the right intraparietal sulcus (47%).

Orthographic transparency and grapheme–phoneme conversion: An ERP study in Arabic and French readers

Numerous behavioral studies have suggested that orthographic transparency of a language is liable to influence the use of grapheme–phoneme conversion during reading. In order to test this hypothesis, the effect of orthographic transparency on event-related potentials was assessed by comparing French to Arab readers. Indeed, French language, contrary to Arabic one, was expected to favor the use of grapheme–phoneme rules during reading. Our results demonstrated that the N320, a component implicated in phonologic transcription, was modulated by orthographic transparency. Indeed, during reading in their mother tongue, only French subjects clearly elicited a N320. Moreover, the comparisons between activations elicited by Arabic words in Arab subjects and French monolingual people also confirm that the N170 component represents an important orthographic stage. The implications of these results on bilinguism and visual word recognition models are discussed.

Altered hemispheric asymmetry during word processing in dyslexic children: an event-related potential study

The present event-related potential study aimed at finding neurophysiological correlates of inadequate reading performances in developmental dyslexia. By using the same set of words in different linguistic tasks, we found variations of cortical asymmetry between dyslexic individuals and controls starting with the N420, a component supposed to index the phonological processing during reading. Whereas in controls this component was left lateralized, in disabled readers it was more distributed across hemispheres. The observed lack of lateralization in dyslexic individuals was also found in the later slow negative wave that developed in the 700 to 1500-ms time interval. We postulate that the altered asymmetry is related to an impairment of the grapheme-phoneme conversion mechanism. The findings would therefore support the hypothesis of a phonological deficit underlying this learning disability.

Reading skills and phonological awareness acquisition in Down syndrome

Although reading abilities play a fundamental role in the acquisition of personal autonomy, up until now studies investigating these abilities in Down syndrome (DS) are aimed at defining educational or rehabilitation acquisition. However, studies describing the relationship between reading and phonological awareness in individuals with DS by comparing them to typically developing children often report contradictory results. The aim of this study is to explore reading and phonological awareness skills in a group of participants with DS. We administered reading and phonological processing ability tests to 17 DS individuals and to 17 reading-age-matched typically developing children. Concerning reading abilities, participants with DS were impaired on non-word reading and on interpreting accuracy of non-homographic homophones. Their passage comprehension was also limited. Comparable ability was reported in the two groups on irregular word reading and passage reading tasks. Regarding phonological awareness ability, individuals with DS showed lower performances on several tasks, such as rhyming, deletion and syllable segmentation. People with DS show particular failure on non-word reading, a task where correct decoding is only partially influenced by lexical access or semantic context. Correct non-word reading mainly requires the use of the grapheme-phoneme conversion process. This process is based on the efficiency of phonological awareness abilities, which are partly impaired in people with DS. The rehabilitative implications of these findings are discussed.

The interaction between semantic and sublexical routes in reading: Converging evidence from Chinese

The classic dual route model for reading (Coltheart, Rastle, Perry, Langdon, & Ziegler, 2001) assumes at least two independent routes from printed input to sound: one via a lexical-semantic pathway and one via a nonlexical pathway which relies on a process of graphemephoneme-conversion (GPC) and leads to an effect of regularity on reading. However, there is evidence from brain-damaged patients that these pathways are not fully independent but instead interact (Hillis & Caramazza, 1995). Although in some patients, impairment prevents either of these routes from independently producing the correct response for a target, the imperfect activation from both routes may summate onto the target leading to a correct response. Weekes Chen, and Yin (1997) reported that oral reading in Chinese can proceed via an independent pathway that bypasses lexical-semantic knowledge. Weekes and Chen (1999) argued this pathway does not support reading of low imageability, low frequency irregular characters resulting in an effect of regularity on reading. This sublexical process on oral reading in Chinese differs from the process of GPC in alphabetic languages. A Chinese character can be considered regular if it contains a visual component (phonetic radical) that can be an independent lexical item with the same pronunciation as the whole character. A character is consistent if all characters containing this phonetic radical have the same pronunciation. Sublexical regularity and consistency play a role in character reading independent of lexical factors in studies with normal (Hue, 1992) and aphasic groups (e.g., Weekes & Chen, 1999). We focus here on the reading of a patient, WJX, who suffers from dementia and who made semantic errors in comprehension and production. His reading ability was well preserved, even for many irregular/inconsistent characters. Our question was what reading process can account for his good reading performance? Case description and results

Toward Spontaneous Speech Synthesis—Utilizing Language Model Information in TTS

State-of-the-art speech synthesis systems achieve a high overall quality. However, synthesized speech still lacks naturalness. To produce more natural and colloquial synthetic speech, our research focuses on integration of effects present in spontaneous speech. Conventional speech synthesis systems do not consider the probability of a word in its context. Recent investigations on corpora of natural speech showed that words that are very likely to occur in a given context are pronounced less accurately and faster than improbable ones. In this paper three approaches are introduced to model this effect found in spontaneous speech. The first algorithm changes the speaking rate directly by shortening or lengthening the syllables of a word depending on the language model probability of that word. Since probable words are not only pronounced faster but also less accurately this approach was extended by selecting appropriate pronunciation variants of a word according to the language model probability. This second algorithm changes the local speaking rate indirectly by controlling the grapheme-phoneme conversion. In a third stage, a pronunciation sequence model was used to select the appropriate variants according to their sequence probability. In listening experiments test participants were asked to rate the synthesized speech in the categories colloquial impression and naturalness. Our approaches achieved a significant improvement in the category colloquial impression. However, no significantly higher naturalness could be observed. The observed effects will be discussed in detail.

The role of left posterior inferior temporal cortex in spelling.

To determine whether damage to left posterior inferior temporal cortex (PITC) is associated with agraphia and to characterize the nature of the spelling impairment. Left angular gyrus may play a critical role in spelling. However, this traditional view is challenged by reports of agraphia after left temporo-occipital lesions and by functional imaging studies demonstrating activation of left PITC during writing in normal individuals. Patients with focal damage to the left temporo-occipital cortex and normal control subjects were administered a comprehensive spelling battery that included regular words, irregular words, and nonwords as stimuli. Although patients performed worse than control subjects in all experimental conditions, the spelling deficit was particularly severe for irregular words, whereas regular word and nonword spelling were less impaired. Additional analyses indicated that orthographic regularity and word frequency had a much more pronounced effect on spelling accuracy in patients compared with control subjects. Most errors on irregular words were phonologically plausible, consistent with reliance on a sublexical phonologic spelling strategy (i.e., phoneme-grapheme conversion). Overall, the spelling impairment of the patients showed the characteristic profile of lexical agraphia. Lesion analyses indicated that the damage in the majority of patients encompassed an area within the left PITC (BA 37/20) where the authors previously obtained evidence of activation in a functional imaging study of writing in normal participants. The behavioral and neuroanatomic observations in the patients are consistent with functional imaging studies of writing in neurologically intact individuals and provide converging evidence for the role of left PITC in spelling. Together, these findings implicate left PITC as a possible neural substrate of the putative orthographic lexicon that contains stored memory representations for the written forms of familiar words.

Electrophysiological correlates of grapheme-phoneme conversion

The cortical processes underlying grapheme-phoneme conversion were investigated by event-related potentials (ERPs). The task consisted of silent reading or vowel-matching of three Japanese hiragana characters, each representing a consonant-vowel syllable. At earlier latencies, typical components of the visual ERP, namely, P1 (110 ms), N1 (170 ms) and P2 (300 ms), were elicited in the temporo-occipital area for both tasks as well as control task (observing the orthographic shapes of three Korean characters). Following these earlier components, two sustained negativities were identified. The earlier sustained negativity, referred here to as SN1, was found in both the silent-reading and vowel-matching task but not in the control task. The scalp distribution of SN1 was over the left occipito-temporal area, with maximum amplitude over O1. The amplitude of SN1 was larger in the vowel-matching task compared to the silent-reading task, consistent with previous reports that ERP amplitude correlates with task difficulty. SN2, the later sustained negativity, was only observed in the vowel-matching task. The scalp distribution of SN2 was over the midsagittal centro-parietal area with maximum amplitude over Cz. Elicitation of SN2 in the vowel-matching task suggested that the vowel-matching task requires a wider range of neural activities exceeding the established conventional area of language processing.

자율 학습을 이용한 선형 정렬 말뭉치 구축

본 논문에서는 자을 선형 정렬 알고리즘을 이용하여 선형 정렬 말뭉치를 구축하는 방법을 제안한다. 기존의 자율 선형 정렬 알고리즘을 이용하여 선형 정렬 말뭉치를 구축할 경우, 두 문자열의 길이가 서로 다르면 정렬된 두 문자열(입력열과 출력열)에 모두 공백문자가 나타난다. 이 방법을 그대로 사용하면 정렬 말뭉치의 구축은 용이하나 정렬된 말뭉치를 이용하는 응용 시스템에서는 탐색 공간이 기하급수적으로 늘어날 뿐 아니라 구축된 정렬 말뭉치는 다양한 기계학습 방법에 두루 사용될 수 없다는 문제가 있다. 본 논문에서는 이들 문제를 최소화하기 위해서 입력열에는 공백문자가 나타나지 않도록 기존의 자을 선형 정렬 알고리즘을 수정하였다. 이 알고리즘을 이용해서 한영 음차 표기 및 복원, 영어 단어의 발음 생성, 영어 발음의 단어 생성, 한국어 형태소 분리 및 복원을 위한 정렬 말뭉치를 구축하였으며, 간단한 실험을 통해, 그들의 실용성을 입증해 보였다. In this paper, we propose a modified unsupervised linear alignment algorithm for building an aligned corpus. The original algorithm inserts null characters into both of two aligned strings (source string and target string), because the two strings are different from each other in length. This can cause some difficulties like the search space explosion for applications using the aligned corpus with null characters and no possibility of applying to several machine learning algorithms. To alleviate these difficulties, we modify the algorithm not to contain null characters in the aligned source strings. We have shown the usability of our approach by applying it to different areas such as Korean-English back-trans literation, English grapheme-phoneme conversion, and Korean morphological analysis.

CHRONOMETRY OF VISUAL WORD RECOGNITION DURING PASSIVE AND LEXICAL DECISION TASKS: AN ERP INVESTIGATION

In order to investigate the neuroanatomical chronometry of word processing, two experiments using: Event Related Potentials (ERPs) have been performed. The first one was designed to test the effects of orthographic, phonologic, and lexical properties of linguistic items on the presemantic components of ERPs during a passive reading task and massive repetition used to reduce familiarity effect between words and nonwords. In a second study, the level of familiarity was investigated by varying stimulus repetition and frequency in a lexical decision task. Overall results suggest a functional discrimination between orthographic and nonorthographic stimuli begun as early as 170 ms (N170 component) whereas the next components (N230 and N320) were sensitive to the orthographic nature of the stimuli, but also to their lexical/phonologic proprieties. The N320 associated to phonological processing (Bentin et al., 1999) was modulated by word frequency and massive repetition caused its disappearance. This suggests that this component may reflect a nonobligatory phonologic stage of graphemephoneme conversion postulated by the DRC model (Coltheart et al., 2001) or semantic phonologically mediated pathway (Harm and Seidenberg, in press).

The size of the cross-lingual masked phonological priming effect does not depend on second language proficiency.

Using a masked phonological priming paradigm, Brysbaert, Van Dyck, and Van de Poel (1999) showed that Dutch-French bilinguals perform better at identifying tachistoscopically presented L2 words (e.g., oui [yes]) when those words are primed by L1 words or nonwords that are homophonic to the L2 target word according to the L1 grapheme-phoneme conversion rules (e.g., wie [who]). They noted that this priming effect was smaller for balanced bilinguals than for less proficient bilinguals, although the interaction failed to reach significance. Findings of Gollan, Forster, and Frost (1997) suggest that this could be attributed to a greater reliance on phonology in L2 reading, caused by a smaller proficiency in this language. However, in this study we show that the Dutch-French cross-lingual phonological priming effect is equally large for perfectly balanced and less proficient bilinguals. Our findings are in line with more recent work of Van Wijnendaele and Brysbaert (2002).

Construction and learning to spell

Little is known about the way in which children learn the rules of literacy. We argue that children’s learning about orthographic rules can be the result of their own constructions. We provide longitudinal evidence on Greek children’s understanding of the morphophonemic rules for the spelling of the /o/ and /e/ vowel sounds in stems and inflections. These sounds can be spelled with more than one phonetically acceptable grapheme. When the vowel sound represents part of an inflection or the inflection itself, the grammatical status of the word is the key to the right choice between the alternative spellings. In contrast, no such morphemic rule applies when the sound is part of the stem morpheme. The spelling of each of these vowel sounds in the stem must rely on rote learning. Our results suggest that children go through a three-step developmental sequence in learning how to spell /o/ and /e/ in inflections: initially they adopt only one spelling for each vowel sound. Later, when they add the alternative spelling to their repertoire, they overgeneralize it to inappropriate words. Eventually, they restrict the alternative spellings to the grammatically appropriate words. We argue that these overgeneralizations, which have also been found in English, French and Portuguese spelling, indicate that children at first try to extend an existing rule, based on grapheme–phoneme conversion, for the new spelling. This slight change produces new experiences for them which allow them to construct a higher-level morphophonemic spelling rule.