Variations in the use of simple and context-sensitive grapheme-phoneme correspondences in English and German developing readers

Abstract

Learning to read in most alphabetic orthographies requires not only the acquisition of simple grapheme-phoneme correspondences (GPCs) but also the acquisition of contextsensitive GPCs, where surrounding letters change a grapheme's pronunciation. We aimed to explore the use and development of simple GPCs (e.g. a ➔ /ae/) and context-sensitive GPCs (e.g. [w]a ➔ / /, as in "swan" or a[l][d] ➔ /o:/, as in "bald") in pseudoword reading. Across three experiments, English-and German-speaking children in grades 2-4 read aloud pseudowords, where vowel graphemes had different pronunciations according to different contexts (e.g. "hact", "wact", "hald"). First, we found that children use context-sensitive GPCs from grade 2 onwards, even when they are not explicitly taught. Second, we used a mathematical optimisation procedure to assess whether children's vowel responses can be described by assuming that they rely on a mix of simple and context-sensitive GPCs. While the approach works well for German adults (Schmalz et al. in Journal of Cognitive Psychology, 26, 831-852, 2014), we found poor model fits for both German-and English-speaking children. Additional analyses using an entropy measure and data from a third experiment showed that children's pseudoword reading responses are variable and likely affected by random noise. We found a decrease in entropy across grade and reading ability across all conditions in both languages. This suggests that GPC knowledge becomes increasingly refined across grades 2-4. Keywords Context-sensitive rules. Cross-linguistic. Entropy. Grapheme-phoneme correspondences. Reading development Most models of single-word reading contain a sublexical route, which converts a string of letters from graphemes to phonemes without relying on whole-word knowledge or semantics

Order among chaos: Cross-linguistic differences and developmental trajectories in pseudoword reading aloud using pronunciation Entropy

“…The Mann-Whitney test showed , by developing and practicing their reading skills children gradually became more acquainted with the GPCs of their language, and their answer variability decreased. This result is in line with the findings of [12], who found decreasing Entropy in vowel pronunciation variability (but not in consonant pronunciation variability, which was not investigated) as a function of grade. In real words, there was no significant difference in lexicalizations between grades(grade 2: m = 0.41, sd = 0.19; grade 3: m = 0.04, sd = 0.21; grade 4: m = 0.03, sd = 0.17): p = 0.594 for the comparison between grade 2 and grade 3; p = 0.524 between grade 3 and grade 4; and p = 0.274 between grade 2 and grade 4).…”

“…Overall, we predicted higher Entropy in younger than in older children, because the knowledge of the GPCs may not be full developed, which could lead to a greater level of noisiness in their decision about how to pronounce a given GPC [12]. Moreover, Entropy was expected to be higher for the English than German items, because the depth of English may make it more difficult for children to learn the GPCs.…”

“…Participants. Six groups of children participated in this experiment: Three groups of monolingual German children, enrolled in in grade 2 (N = 22), grade 3 (N = 19), and grade 4 (N = 22) (for a more detailed description of this sample, see [12]) were recruited in German primary schools, as well as, three groups of German/English bilingual children attending grade 2 (N = 12), grade 3 (N = 5), and grade 4 (N = 5) of a bilingual primary school in Australia. Prior to testing informed consent was obtained from children's parents.…”

“…Entropy is a concept first introduced by Shannon's Information Theory [13], which can be defined as the degree of chaos within a closed system. Earlier studies in psycholinguistic research used Entropy as a measure to investigate processing difficulty in sentence comprehension [14], quantify orthographic transparency in different orthographies (using word onsets: [15][16][17], using mono-syllabic words: [18], using whole words: [19]), and to assess variability in responses to disyllabic English pseudowords [20] as well as diversity in vowel pronunciation in German and English children reading aloud pseudowords [12].…”

“…Our study's goal was to evaluate the use of Entropy (H) in participants' pseudoword reading aloud responses. Although Entropy has already been used to measure the diversity of vowel pronunciations in German and English children reading aloud pseudowords across grades [12], alternative pronunciations of disyllabic pseudowords in English [20], we are the first, to our knowledge, to use it to compare individual responses to both mono-syllabic and multi-syllabic pseudowords across age (primary school children and adults) and languages (shallow and deep orthographies), including a consideration for bilingualism (in children).…”