Syllable-first rather than letter-first to improve phonemic awareness

The present study investigates the nature of the spelling-to-sound correspondences taught to enhance phonemic awareness in prereaders. The main assumption in the literature is that learning the alphabetic code through letter-to-phoneme correspondences is the best way to improve phonemic awareness. The alternative syllabic bridge hypothesis, based on the saliency and early availability of syllables, assumes that learning to associate letters to phonological syllables enables phoneme units to be the mirror of the letters and to become accessible, thereby developing phonemic awareness of prereaders. A total of 222 French-speaking prereaders took part in a 4-session learning program based on correspondences either between letters and syllables (letters-to-syllable group) or between letters and phonemes (letter-to-phoneme group), and the fifth last session on coding and decoding. Our results showed a greater increase in phonemic awareness in the letters-to-syllable group than in the letter-to-phoneme group. The present study suggests that teaching prereaders letters-to-syllable correspondences is a key to successful reading.


Results
The aim of the first analysis was to test the progression of children's skills in reading syllables between T1 and T3. The second analysis aimed to test our hypothesis by testing the progression of phonemic awareness skills in each group. The third analysis was conducted on the progression of phonemic awareness skills as a function of children's initial skills at T1.
Progression of syllable reading. All children progress significantly between T1 and T3 on syllable reading (F(1,220) = 212.79, p < .0001). Moreover, syllable reading performance (Fig. 1B) increased between T1 and T3 more in the letters-to-syllable group (1.4% vs. 41.6% of correct responses) than in the letter-to-phoneme group (1.4-22.3% of correct responses; F(1,220) = 21.25, p < .0001. However, this result cannot be taken as a solid evidence of the better efficiency of the letters-to-syllable training over the letter-to-phoneme training as the letters-to-syllable group was specifically trained to read syllables in the first four sessions. Then these data should simply show the efficiency of the first four teaching sessions. Similarly, knowledge of letter names and letter sounds increased between T1 and T3 more in the letter-to-phoneme group (58.4% vs. 85.1% of correct responses) than in the letters-to-syllable teaching group (58.8% vs. 68.9%; F(1,220) = 60.21, p < .0001, Fig. 1A).
Development of phonemic awareness. The main effect of time (T1, T2, T3) was significant on phoneme awareness scores, (F(2,440) = 162.18, p < .0001), but not the main effect of group, (F(1,220) = 1.48, p = . 22). The main result was in agreement with our main hypothesis, phoneme awareness skills increased more in the letters-to-syllable group (T1: 30.4%; T2: 66.3%; T3: 74.6% of accuracy) than in the letter-to-phoneme group (T1: 32.9%; T2: 56%; T3: 64.8% of accuracy), as shown in Fig. 2 the Time x Group interaction was significant, F(2,440) = 5.31, p = .005. Planned comparisons (Bonferroni corrected) revealed a significant interaction between group and time when T1 and T2 were selected (F(1,220) = 6.72, p = .01), but not with T2 and T3 (F(1,220) = 0.02, p = .87). www.nature.com/scientificreports/ A second concern was whether children were making progress on the final phoneme elision task only on items on which they had received instruction (i.e., learned syllables), or whether they transferred their skills to other items (i.e., syllables that were not learned and new syllables, including new letters), see Table 1. No significant interaction was found between group, time and syllable items, F(4,880) = 0.92, p = .44). The progress in phoneme awareness in the letters-to-syllable group compared to the letter-to-phoneme group was significant for the learned syllables (F(1,220) = 7.29, p = .007) and for new syllables (F(1,220) = 7.30, p = .007), and marginal for syllables that were not learned (F(1,220) = 3.60, p = .05).
Further investigation of developmental phonemic awareness as a function of the child's initial knowledge. While the effect of training on phonemic awareness appeared to be greatest in the group of children who were taught letters-to-syllable correspondences, we also explored whether the effectiveness of the training was influenced by the children's initial knowledge. As there was marked inter-individual variability in the measurements made at T1 (see Fig. 3), we divided the cohort of 222 children for analysis into four subgroups based on their initial level of phonemic awareness and letter-name knowledge (i.e., the two most important predictors of future reading ability). Children were classified as having good skills if their score was above average and as having poor skills if their score was below average ( Table 2).
We did not analyze the subgroup with initial high phoneme awareness and low letter name knowledge because there were only 11 participants in this subgroup (n = 11).

Discussion
Previous studies showed that phonemic awareness develops with reading instructions on print-to-speech mapping. The present study used phonemic awareness to capture the very first signs of the mastery of the alphabetic code. It aimed to test whether the development of phonemic awareness-as a marker of the alphabetic code acquisition-was enhanced when phonics instruction focused on letters-to-syllable or on letter-to-phoneme correspondences. The first analysis showed that the letters-to-syllable group progress better in syllable reading than the letter-to-phoneme group between T1 and T3. These data have been expected as we trained the lettersto-syllable group to read syllables during the teaching program. Our main result was the interaction between time and teaching group: the teaching of letters-to-syllable relationships was more conducive to the development of phonemic awareness than that of letter-to-phoneme relationships. More precisely, the greater progress made by the letters-to-syllable teaching group than by the letter-to-phoneme group was observed between T1 and T2,  T1  T2  T3  T1  T2  T3  T1  T2  T3 Percentage of correct responses in phonemic awareness task L-L groups L-H groups H-H groups Letter-to-phoneme training group Letters-to-syllable training group www.nature.com/scientificreports/ but not between T2 and T3. This result is evidence that the greater progress in phonemic awareness was due to learning letters-to-sound correspondences mediated by syllable units. Thus, the success of this learning was due to its associative nature and not to explicit instructions concerning the alphabetic principle. In contrast, a single introduction session to coding and decoding did not boost progression in phonemic awareness, nor did it alter it, since the advantage of the syllable teaching group was maintained at T3. However, we cannot exclude the possibility that this was due to the brevity of the coding and decoding session. The main difficulty facing children who are learning to read is that, independent of each other, phonemes have no physical reality. That is why teaching methods promote learning of correspondences directly between the letter and the phoneme. Contrary to the dominant opinion that the attentional focus must be directed towards phonemes and their relations to letters, here we show that focusing on syllable units is more effective in developing phonemic awareness, as proposed in the syllabic bridge hypothesis 103 . As noted above, syllables are the easiest and earliest units to perceive in spoken language and to manipulate for prereaders. Our concept is the following: by teaching the relations between letters and syllables, children learn the pronunciation of letters depending on their context, and learn that a spoken syllable corresponds to a sequence of letters; this implicit and associative learning allows children to extract regularities between letters and phonemes. Then they can build phoneme representations in the mirror of letter representations.
This striking result has important implications for enhancing the process of learning to read. Phoneme awareness is the ability the most strongly correlated with reading acquisition. Its facilitative effect is strongest during the period in which children learn to crack the alphabetic code 8 . Phonemic awareness is thus considered as a marker of the alphabetic code acquisition in the learning-to-read process 12,19,52 . The present study clearly shows that automatic connections between letters and syllables boost the acquisition of phonemic awareness and therefore mediate successful reading acquisition. It is important to note that the children were not previously trained in phonemic awareness, thus proving that the teaching of simple relationships between letters and syllables leads to progress in phonemic awareness skills.
The complementary analysis of the data concerning the children's initial skills has two major interests. On the one hand, it describes the population of prereaders according to the two factors that influence the development of skilled reading. The majority of prereaders (49.5%) had good knowledge of letter names, but low phonemic awareness skills and 29.2% already had both skills. In contrast, 16.2% had not yet developed either of the skills, and only 4.9% had high performances in phoneme awareness but low knowledge of letter name. Even if phonemic awareness and letter name knowledge influence one another as they develop 7,25 , our analysis of 222 prereaders clearly showed that they develop more skills in letter name than in phonemic awareness before formal reading instruction begins. This analysis leads us to conclude that there is a developmental progression of knowledge before learning to read: a) children do not present any phoneme awareness and letter name skills; b) children begin to learn the letter names; c) then children begin to develop phoneme awareness.
On the other hand, the complementary analysis provides information on how prereaders learn from the instructions they receive. The prereaders, for whom teaching the letters-to-syllable relationships was more conducive to phonemic awareness progression than that of letter-to-phoneme relationships, were prereaders with high initial knowledge of letter names (with low or high phonemic awareness, representing three-quarters of the sample). In contrast, no difference was observed between the two types of teaching on the development of phonemic awareness in prereaders with limited knowledge of letter names. This result leads us to assume that knowledge of letter names is a prerequisite for benefitting from phonics instruction based on syllable units. As Adams 51 recalls, new knowledge is built on existing knowledge. Hence, children first develop knowledge of letter names as they "provide a more accessible link between print and speech than do letter sounds" 7 p. 595, and thereafter they can learn to associate letters with available phonological syllable units and consequently develop their phoneme awareness.
In conclusion, our results shed doubt on the widely accepted idea that children need to learn the correspondences between letters and phonemes in order to develop phoneme awareness. Instead, children should learn the correspondences between letters and the larger accessible phonological units, syllables. However, the acquisition of a syllabary is clearly not envisaged. Rather, we propose that after having acquired a basic knowledge of letter names, prereaders could benefit from learning the letters-to-syllable relationships that make sense given their prior knowledge and skills. From a bundle of letters-to-syllable connections, prereaders could build phoneme representations in mirror of letters, and thereafter acquire and master the alphabetic code.

Method
Participants. A total of 415 children attending public preschool participated in our study, and 193 were excluded. First, 65 children were excluded because none of the parents at home spoke French; then, 46 children were excluded because they were already readers at T1; finally, 82 children were excluded because they were absent from at least one teaching session or one of the three tests. The final sample comprised 222 French children (of which 136 girls), aged 5 years and 4 months on average (Table 3), from 15 different preschools. More precisely, 160 children attended in 11 schools with medium to high socioeconomic status and 62 children attended in 4 schools with low socioeconomic status, equally distributed between the two groups. They all had normal vision and hearing and presented no language disorders. Participants in both groups were matched using three literacy scores at T1 (Table 4). Child's parents or legal guardians provided informed consent prior to inclusion in the study in accordance with the Declaration of Helsinki. This study was approved by the ethics committee of the Grenoble-Alpes University (No. IRB00010290-2017-12-12-34) and by the Director of departmental services of the Ministry of National Education ("DASEN"). All methods were performed in accordance with the aforementioned relevant guidelines and regulations. Scientific Reports | (2020) 10:22130 | https://doi.org/10.1038/s41598-020-79240-y www.nature.com/scientificreports/ Procedure and material. The design of the experiment (Fig. 5) was the following: T1 (i.e., pre-test)teaching (four sessions)-T2 (i.e., post-test 1)-coding-decoding introduction (one session)-T3 (i.e., post-test 2). All characters (syllables and letters) were printed in capitals Calibri 72, using black type on a sheet of white A4 paper. The same typography was used for the material used for the teaching sessions and the tests.
Learning program. Children in both the letters-to-syllable group and letter-to-phoneme group participated in four sessions of associative learning and one introductory session to coding and decoding. Each session lasted 25 min (the usual duration of learning activities in French kindergartens) distributed over a period of 4 weeks (total learning time 125 min) in order to enhance learning 125 . Each session was conducted in small groups of 4-5 children. The letter-to-phoneme group followed a control program built with language exercises usually taught in French preschools based on learning of correspondences between a grapheme and a phoneme. Four consonants (C) with only one phonemic identity in onset position in French and 4 vowels (V) were chosen ("b", "f ", "t", "s", "a", "i", "o", "u"). The letters-to-syllable group followed a program centered on letters-to-syllable learning. Syllables were constructed with the same consonants and vowels as in the letter-to-phoneme group; all the syllables were CV structure, which is the most frequent in French. In a Latin-square design and in order to avoid effects of syllable material, two subsets of 8 syllables were created ("ba", "bi", "fa", "fi", "so", "su", "to", "tu" and "bo", "bu", "fo", "fu", "sa", "si", "ta", "ti"); randomly, half the children in the letters-to-syllable group learned set 1, and the other half learned set 2 (see S1 Supplementary information).
Teaching sessions. The aim was to teach children correspondences between written and spoken syllables (in letters-to-syllable group) and between letters and the corresponding name and sound (in the letter-to-phoneme group). In the first four sessions of the program, the children learned a total of 8 syllables or 8 letters depending on which group they were in; each syllable or letter being taught in at least two sessions. Five exercises were  www.nature.com/scientificreports/ proposed during the sessions: reading, letter-phoneme matching (only in the letter-to-phoneme group), dice, lotto and relay exercises. In the reading exercise, children were asked to read three times; 1) a syllable or a letter that was presented visually while the instructor read the syllable or give the name and the sound of the letter, the children then had to repeat the syllable or the letter name and sound all together, and so on, for the three other syllables or letters; 2) the children had to read the four syllables aloud together or give the name and sound of the four letters, without the instructor; 3) each individual child had to read one of the four syllables or give the name and sound of the four letters. The letter-phoneme exercise was only added in the letter-to-phoneme group (in the two first sessions of the program) so that the duration of the teaching program was the same in the two groups. In this exercise, children had to find the first letter in the name of four pictures (e.g., the letter "s" in the picture of the "soleil", sun in English). In the dice exercise, children had to read the syllable aloud or give the name of the letter and the sound of the syllable or the letter written on the upper side of a rolled dice; when the child gave the correct response, he/she moved his pawn forward on a 3-square board. In the lotto exercise, each child received a lotto grid with four written syllables or four letters; they had to listen to the sound of the syllable or the name of the letter pronounced by the instructor and locate the corresponding written syllable or letter on the lotto grid.
In the relay exercise, which was played by a team comprising two children, the first child had to read a writtensyllable aloud or produce a letter name and sound; the second child had to listen and memorize the phonological syllable or letter name and sound just heard, cross the classroom and, in a set of four written syllables or four letters, find the matching syllable or letter card, and check back with the first child; they then changed roles. The fifth session of the teaching program was an introductory session on coding and decoding and was the same in both groups. The coding exercise corresponds to what Castles et al. (2018) call the analytic phonic programs, which "begins with whole words, and [for which] grapheme-phoneme correspondences are taught by breaking those words down into their component parts." (p. 13): children had to choose the individual letters that corresponded to the syllable pronounced by the instructor (e.g., the instructor said /bo/ and the instruction given to children was to find the letters that correspond to the syllable; in case of failure, the correct answer was immediately given). The decoding exercise corresponds to synthetic phonic programs which "teach grapheme-phoneme correspondences individually and in a specified sequence, and children are taught early to blend (synthesize, hence the term synthetic) individual phonemes together to make words" (Castles et al., 2018, p. 13): children had to pronounce the syllable composed by the individual letters chosen by the instructor (e.g., the instructor gave the letter "b", which sounds /b/ and the letter "o", and the instruction given to children was to find the correct pronunciation of the two letters put together; in case of failure, the correct answer was immediately given). To respect the order of the teaching program, the two exercises were not presented to the two groups in the same order. As the letter-to-phoneme group focused on letters and phonemes, they started with the coding exercise in which they had to combine two letters. In contrast, as the letters-to-syllable focused on letters and phonological syllables, they started with the decoding exercise in which they had to break the syllable into phonemes and letters.
Tests. Three measures of early literacy were used: letter knowledge, phonological awareness, and syllable reading. All three tasks were measured at T1 and T3, whereas only phonological awareness was measured at T2. Task 1: Letter knowledge. We used the same material as that used in the teaching sessions (i.e., "a", "i", "o" and "u"; "b", "t", "f " and "s"). Children had to say the name of each vowel and consonant and the sound of the consonants aloud. No feedback and no stopping rule when children failed were given. Task 2: Final phoneme elision. The examiners presented a stimulus syllable with a CVC structure verbally and asked the children to produce the target syllable by repeating it without the final phoneme (e.g., /bak/ = > / ba/). The examiner provided one example and two trials with feedback before starting the test. During the test itself, no feedback was given. Among the 24 items, eight were syllables learned in the teaching session (i.e., set 1 or set 2), eight were syllables that had not been learned (i.e., set 2 for participants who learned set 1; set 1 for participants who learned set 2), and eight were new syllables composed of new consonants that had not been learned in the program, such as "v" /v/, "p" /p/, "m" /m/, "r" /R/ (e.g., "vip" /vip/, "mul" /myl/). A stopping rule was applied after four incorrect items for learned syllables, and after two incorrect items for unlearned syllables and new syllables. Task 3: Syllable reading. Children were asked to read 16 CV syllables, eight learned syllables and eight unlearned syllables aloud. No feedback and no stopping rule were given. Table 3 details the program and lists the material used in each group (i.e., letters-to-syllable teaching program and letter-to-phoneme teaching program).
Statistical analysis. All analyses were performed using STATISTICA software. Although the repeated measures were not normally distributed, it is generally accepted that ANOVA is a robust test, particularly when the sample is large enough and the groups are about the same size [126][127][128] . An ANOVA with group (letters-tosyllable vs. letter-to-phoneme) as a between-subject variable, time (T1, T2 and T3) and syllables (learned, not learned and novel syllables) as within-subject variable was conducted on measures of phoneme awareness. In the case of planned comparisons, the Bonferroni method was used and the significance level has been adjusted (to 0.05/2 = 0.025). ANOVAs with group and time (only T1 and T3) were conducted on letter knowledge and syllable reading scores. The datasets generated during and analysed during the current study are available in the Open Science Framework repository, https ://mfr.osf.io/rende r?url=https %3A%2F%2Fosf .io%2Febp ft%2Fdow nload .