3.1.1. Participants, teachers and learning context

A total of 75 ELLs from two public primary schools in Lower Saxony, Germany, took part in this longitudinal, quasi-experimental study. In Germany, children typically begin their formal education with primary school, which starts at the age of six. The participants were divided into three classes, each assigned randomly to one of the three literacy intervention groups: Phonics, whole word and combined. They were followed from grade 3 to grade 4. Before the study began, all the children had learned fundamental literacy skills in German.

In the phonics group, there were 26 children (13 females, 13 males) with an average age at the end of the intervention (T2) of 10.5 years (SD = 0.76), ranging from 9.72 to 11.76 years. The whole word group consisted of 12 girls and 11 boys with a mean age of 10.57 years (SD = 0.73), spanning from 9.94 to 11.99 years. The combined group included 26 participants with an average age at T2 of 10.72 years (SD = 0.60), with ages ranging from 9.98 to 11.84 years. ELLs were predominantly native German speakers (21 in phonics, 19 in whole word and 21 in combined). Children within these groups also spoke languages including Arabic, Bulgarian, Kurdish, Polish, Russian, Hungarian and Turkish as their L1, with German as their L2. Questionnaire data indicated that all non-native German-speaking participants were either born in Germany or had relocated before the start of formal education, allowing for early familiarisation with the Latin alphabet. Additionally, those children with non-Latin script backgrounds reported they were orally fluent but not literate in their first languages. Among the participant groups, the phonics group contained a child with dyslexia, the whole word group had two children with distinct learning disabilities (general learning difficulties and dyscalculia) and the combined group had a child with social-emotional challenges. Our study encompasses a wide range of learners, thereby reflecting classroom diversity in Germany, which aligns with the country’s inclusive educational policies. In addition, we used parental educational level and occupation to measure family socioeconomic status (SES). Table 1 summarises participants’ demographic information.

The EFL instruction for the three classes was delivered by two experienced teachers, both holding university degrees in teaching English as a foreign language and each with substantial experience in the field (35 and 26 years of teaching experience, respectively). In our study, we established the comparability of both the instructional conditions and English language exposure across the three experimental groups. Each group received EFL instruction twice a week, with each lesson lasting 45 minutes. Over the course of the study, the total instruction time amounted to 76 weeks for the group receiving phonics-based instruction, 77 weeks for the group focused on the whole word approach and 75 weeks for the combined method group. Nonetheless, these differences were minimal and did not impact the overall comparability of the ‘dosage’ of English language input and literacy practice received by students in each group. All groups were provided with the same core instructional materials, which were thoughtfully adapted to ensure that each set of materials comprehensively supported the distinct teaching methods of the respective intervention groups. Parent questionnaires revealed that exposure to the English language had been very limited for all learners before entering grade 3. Contact time with English was mostly established through L2 teacher input, auditory learning materials, EFL textbooks and other written learning materials such as children’s books and worksheets.

Intervention logs provided a means to track implementation fidelity. This was controlled for by additionally using the Teacher Input Observation Scheme (Kersten et al., 2018) for 10 videotaped lesson recordings in each group to compare the number of cognitively stimulating activities, verbal input, non-verbal input and the support of learners’ output across groups. Three trained students, who had undergone extensive training on the strategies, curricula and observational protocols of the interventions, conducted fidelity checks. These assistants were not directly involved in the implementation of the intervention. Their examination focused on assessing adherence to the intervention, frequency and duration of exposure, quality of delivery, participant responsiveness and programme differentiation.

3.1.2. Literacy intervention programmes (phonics, whole word and combined)

The literacy component of the intervention programmes focused on either phonics, whole word or combined instruction. We included the whole word approach in this study as 74% of German primary teachers reported using this method in the primary EFL classroom (Treutlein et al., 2013) in keeping with market-leading coursebooks.

Literacy programme A (phonics instruction) focused on the explicit teaching of the systematic relationships between letters and sounds in English as an alphabetic writing system. The key concepts were increasing children’s knowledge of English GPC/PGC and pointing to the differences between the two orthographies (German and English). This should have helped the children to learn to spell familiar and unfamiliar words, without them encoding the whole written form using L1 spelling strategies or attempting to retrieve the word form from memory. Regular high-frequency words were fully sounded out phonetically. For exception words, often referred to as “tricky words”, the children learned to identify and decode/encode the parts of the word that followed phonetic rules and to sound out/spell as much of the word as they could. In the phonics programme, PGC that are the same in English and German, such as /t/, /k/ and /m/, were not explicitly taught. Special attention was given to PGC that differ between the two orthographies, for example /j/ as in ‘jump’, ‘jam’ and ‘enjoy’; and the /oo/ sound in ‘food’, ‘room’ and ‘too’.

Literacy programme B (whole word instruction) targeted the lexical reading and spelling route through sight word training (i.e., recognising whole words from orthographic memory). This instruction focused on holistic word recognition, in which lexical items were learned, stored and retrieved by memorising their shape. In this teaching method, often also called sight word or ‘look and say’ instruction, individual words were not broken down into phonemes or other sub-lexical parts but were used as whole units. The focus was on the word as the minimum unit of meaning. In the classroom, words were shown to students in various demonstrations and forms provided by the two teachers (e.g., in different sentences and pictures). Learners built a large sight word vocabulary, labelled things in the classroom, used picture dictionaries, sorted whole sentences to retell stories, used visualisation techniques and played memory games to boost their visual memory. Traditional drill and practice was used as a flashcard technique.

Programme C (combined instruction) differed from programme A (phonics) in two distinct areas: The quantity of letter-sound mappings and the methodology employed for instructing high-frequency words. In programme C, children were taught to use two strategies when reading and spelling. For reading, they could either recognise the word by sight or decode it phonetically. However, for spelling, children were taught to either recall the spelling of high-frequency words from memory or to apply phonetic segmentation and recombine the sounds to spell the word accurately. This combined method contrasts with programme A, where children were exclusively taught phonic decoding and encoding. Literacy programme C targeted teaching a combination of phonics skills together with whole word training. Children were provided with phonics instruction that emphasised the most common pronunciation of each grapheme. Subsequently, they were introduced to a selected list of 50 sight words, which comprised both regular words (e.g., ‘had’, ‘up’ or ‘went’) and exception words (e.g., ‘have’, ‘does’ or ‘one’). The aim of the instruction was to facilitate children in the recognition and retrieval of each of these words as whole entities, rather than through the decoding of individual components. In contrast, programme A employed an alternative approach. In this case, participants were given explicit instructions to fully sound out regular words, systematically implementing phonics principles. When confronted with exception words, the methodology involved recognising phonically decodable segments within the words and decoding as much of the words as the rules permitted.

3.1.3. Data collection and procedure

Data collection protocols for our study were designed to incorporate literacy, cognitive and linguistic measures. Cognitive measures were administered as pre-test control variables at the beginning of grade 3, while linguistic variables were assessed both at the start and then again at the end of grade 4 to evaluate progress following the interventions. Testing was conducted in different settings to accommodate the nature of each task. Participants were tested individually in a quiet room for tasks measuring L2 oral reading rate, L2 oral reading accuracy, phonological short-term memory, working memory and phonological awareness. Group assessments for L2 reading comprehension, L2 real word and pseudoword spelling, L2 grammar/vocabulary and non-verbal intelligence were carried out in the classroom setting. For all measures, instructions were consistently given in German to ensure comprehension across the participant pool. To maintain the standardisation of data-collection processes, the order of test administration was identical for all participants. Prior to conducting the tests, written parental consent was obtained in accordance with the research ethics guidelines set by the school authority. The data-collection team comprised researchers and graduate students who had received thorough formal training in the administration and scoring of the specific measures used. Oral reading rate and accuracy were evaluated by two English native speaker raters, yielding a Cohen’s Kappa of 0.83.

3.1.4 Assessment measures

To examine the lexical and sub-lexical spelling abilities of the participants, we administered both a real-word and a pseudoword spelling task. The real-word spelling test aimed to evaluate the children’s ability to retrieve word-specific spellings from memory, a skill which relies on the lexical route often used for familiar and irregularly spelled words. In contrast, the pseudoword spelling test was designed to assess participants’ proficiency in applying PGC rules in the absence of established lexical representations, a critical aspect of encoding unfamiliar words. By employing pseudowords that require spellers to rely on phonological processing, we aimed to shed light on how the sub-lexical route supports the development of orthographic skills when encountering novel orthographic forms.

In the L2 real-world spelling task, we measured German L1 speakers’ ability to handle both familiar (non-novel) and unfamiliar (novel) phonological and orthographic properties of English. The assessment tasks were divided into four categories. The first category targeted the spelling of eight words containing four phonemes that are novel for German L1 speakers (/æ/, /θ/, /ð/ and /ɒ/). The second category focused on evaluating the spelling of four words with orthographic features that are unfamiliar in German orthography, such as the silent <e> (e.g., ‘like’) and /ʃ/ spelled <sh>. The third category covered seven high-frequency words, including words with regular (e.g., ‘three’) and irregular PGC (e.g., ‘does’). The fourth category consisted of four words with two non-novel phonemes, /e/ and initial /h/. Participants were presented with 23 British English words, each articulated three times, twice in isolation and once within a sentence. Learners then transcribed these words onto paper.

For the L2 pseudoword spelling task, we used the same words from categories I, II and IV of the real word spelling tasks to create pseudowords (e.g., ‘shof’ for ‘shop’ or ‘thip’ for ‘this’). We carefully restricted the number of orthographically plausible responses to ensure validity of the findings. Participants listened to each of the 16 pseudowords, pronounced twice in British English, then wrote down what they heard.

Turning to reading measures, L2 reading comprehension was assessed using two different tests. Test A (Halatschev et al., 2018) is a standardised test specifically designed for 4th-grade ELLs in Germany. This test involved silently reading a passage and answering ten multiple-choice questions. Test B, based on Otto, the Little Spider (van Genechten, 2015), consisted of 19 questions of mixed format, including multiple-choice, true/false, picture sequencing and sentence completion. Both assessments were administered in a paper-and-pencil format.

L2 oral reading rate and accuracy were assessed by having participants read aloud from the children’s book Pirate Pat (Mackinnon et al., 2010). The text was presented in plain format, devoid of any visual cues from the original publication, to ensure that the assessment focused solely on reading ability. Each participant read the text individually, and the performance was video-recorded for later analysis. The read-aloud test had no time constraints, allowing participants to read at their natural pace.

There were four cognitive measures pertaining to working memory, phonological short-term memory, non-verbal intelligence and phonological awareness. Children’s working memory was evaluated with the Wechsler Intelligence Scale for Children–4th Edition’s Digit Span Backwards (DSB) and Letter-Number Sequencing (LNS) subtests, with DSB testing reversed digit recall and LNS testing letter and number arrangement and recall.

For phonological short-term memory, the measurement of verbal immediate memory was conducted using the Digit Span Forward (DSF) sub-test from the Wechsler Intelligence Scale for Children–4th Edition (Petermann & Petermann, 2011), where learners repeated a series of numbers in the same order as they were presented on a sound recording.

Participants’ non-verbal reasoning abilities were evaluated by using a sub-test from the German school readiness test (Esser et al., 2008), in which incomplete matrices had to be completed by selecting the correct element from a range of five to eight alternatives.

Finally, the Test of Basic Competencies for Reading and Spelling Skills (Stock et al., 2017) was employed to evaluate learners’ phonological awareness at a phoneme level. Three sub-tests were chosen: Non-word segmentation, vowel substitution and vowel length.

We also employed two L2 linguistic measures. Participants’ English receptive grammar was assessed through the Early Language and Intercultural Acquisition Studies Grammar Test II (Kersten et al., 2012), a comprehensive evaluation of 12 grammatical phenomena, such as possessive pronouns, negation or passive. A recent study by Koch et al. (2021) confirmed the test’s validity using a Rasch model and demonstrated a strong model fit in a bootstrap goodness-of-fit test. Second, L2 English receptive vocabulary was assessed through the norm-referenced British Picture Vocabulary Scale III (Dunn et al., 2009). The test was administered in a group setting using a recorded format. Participants each had their own paper-based test booklet, which presented four images per test item across sets A to J. Children were instructed to listen to each word played from the recording and then to mark the corresponding image in their booklet. In a study with 388 primary school children (Ponto, 2024), the internal consistency of the dataset for the British Picture Vocabulary Scale III was found to be high (α = 0.94).

3.1.5. Spelling error analysis

Spelling errors were categorised using a modified version of the Phonological, Orthographic and Morphological Assessment of Spelling (POMAS) framework (Bahr et al., 2012), which draws on the Triple Word Form Theory as a linguistically-informed, qualitative scoring system (e.g., Berninger et al., 2010). The POMAS coding scheme was adapted to focus exclusively on phonological and orthographic errors. Morphological errors were not included in the analysis as the words in the study sample did not involve the process of morphological word formation. As the learners in our study were emergent L2 spellers, we adapted the POMAS spelling error analysis to code the errors into clear categories (phonological or orthographic) depending on whether ELLs violated primarily phonological (affecting the phonological plausibility of a word) or orthographic (phonologically plausible but lack of orthographic conventions) rules and regulations. Within our sample, phonological errors were defined as those influencing the phonetic representation of words. Examples included substitutions of similar-sounding phonemes, such as ‘wis’ for ‘with’ (<s> for /ð/), ‘fin’ for ‘thin’ (<f> for /θ/), ‘sauf’ for ‘south’ (<f> for /θ/), ‘swie’ for ‘three’ (<s> and <w> for /θ/ and /r/), or ‘dis’ for ‘this’ (<d> for /ð/). Orthographic errors in our sample typically involved the incorrect application of spelling patterns to represent sounds. For instance, learners might spell ‘does’ as ‘das’, ‘shop’ as ‘schop’, ‘want’ as ‘onet’, ‘had’ as ‘hat’ or ‘around’ as ‘eround’. These mistakes indicate that the children may not have fully acquired the complex spelling patterns of English or understood when to apply them. Next, we classified the errors in the sample on a more fine-grained level, using a more detailed classification of errors according to specific linguistic features derived from general British English. Specifically, we were looking for interference from L1 German for novel phonological (/æ/, /θ/, /ð/, /ɒ/) and orthographic (silent <e> and /ʃ/ spelled as <sh>) patterns. These errors are representative of language-specific differences between German and English. The sample was scored by one of the authors and rescored by two trained research group students. The inter-rater reliability, as analysed by Cohen’s Kappa coefficient, was 0.86 (almost perfect agreement).

3.2.1. Group differences

This study aimed to compare L2 spelling, reading and linguistic abilities in young L1 German ELLs following two years of phonics, whole word or combined instruction. We assessed differences in these abilities at T2 (end of grade 4) using a univariate analysis of variance. The results, including the total number of phonological and orthographic errors, are presented in Table 3. The instructional method was treated as a between-subjects factor. At T2, significant differences were observed in L2 reading rate (p = .048) and phonological errors in spelling (p = .043) across the groups. No other significant differences were noted. These findings indicate that combined instruction seemed to enhance children’s L2 reading rate and that children in the phonics group committed the fewest phonological errors in spelling.

To determine which groups differed significantly for the variables with significant overall tests (i.e., L2 reading rate and L2 phonological errors), we conducted Bonferroni-corrected post-hoc tests. The p-values reported for these post-hoc tests were adjusted in accordance with the Bonferroni correction method as implemented in SPSS, where the observed empirical p-value is multiplied by the number of tests conducted. For L2 reading rate, the combined group (M = 79.77 words per minute) showed a higher mean reading rate compared to the whole word group (M = 68.09 words per minute), with this difference approaching significance (p = .055). Regarding L2 phonological errors, the whole word group (M = 11.00, SD = 3.08) had a significantly higher mean number of errors than the phonics group (M = 7.88, SD = 4.57), with a p-value of .037.

We further examined the development of children’s receptive L2 grammar and L2 vocabulary knowledge across the three instructional groups. For both L2 grammar and L2 vocabulary measures, which were available at the beginning of grade 3 (pre-intervention) and at the end of grade 4 (post-intervention), we conducted a two-factorial ANOVA with measurement time as a within-subject factor and instructional method (phonics, whole word and combined) as a between-subject factor. This analysis allowed us to assess the interaction effects between time and instructional method on the learners’ language development. The results are presented in Table 4.

For L2 grammar, significant improvement over time was observed (F(1, 72) = 67.95, p < .001), with no significant differences between the instructional conditions (F(2, 72) = 1.45, p = .242) or their interaction with time (F(2, 72) = 2.46, p = .093). These findings suggest that while all groups improved from the beginning to the end of the study, the rate of improvement was consistent across different instructional methods.

For L2 vocabulary, we found significant main effects for time (F(1, 72) = 190.78, p < .001) and a significant interaction (F(2, 72) = 25.65, p < .001) but no main effect for the condition (F(2, 72) = 2.43, p = .096). To specify the meaning of the significant interaction, we calculated conditional main effects. When comparing the average values between T1 and T2 for all three groups, we found significant conditional main effects for all three groups (all p < .001). For T1, we found no significant group differences for L2 vocabulary (F(2, 72) = 2.73, p = .072). For T2, however, we found a significant conditional main effect (F(2, 72) = 10.82, p < .001). Bonferroni-corrected post hoc tests indicated that the phonics group showed significantly higher values compared to both other groups (both p < .001) while the whole word and combined group did not differ at all (p > .05) regarding their L2 vocabulary. Whereas all three groups did not differ regarding their L2 vocabulary knowledge at the onset of the study, their receptive lexicon size increased with varying outcomes, resulting in a significantly higher average value in the phonics group compared to both other groups.

3.2.2. Error analysis: Phonological and orthographic errors

Further, we investigated the extent to which young German ELLs used phonological and orthographic knowledge when spelling in English and the impact of a two-year instructional programme on this process.

Out of the 75 children who spelled a total of 1725 real words, 911 errors were found, accounting for 52.81% of the words produced. Errors were found in both POMAS categories we used for error analysis, phonological and orthographic. Specifically, 701 errors were phonological in nature, representing 76.95% of all errors, while 210 errors were due to orthographic misspellings, making up 23.05% of the total. Analysing the distribution of errors across groups, it is evident that the prevalence of error types varied (see Figure 1). We found significant group differences between the phonics and whole word group, F(2, 72) = 3.29, p = .043. The fewest phonological errors occurred in the phonics group (M = 7.88, SD = 4.57), followed by the combined (M = 9.35, SD = 4.75) and whole word group (M = 11.00, SD = 3.08). With regard to the distribution of orthographic errors, there were no significant differences between groups. Again, fewest errors were made in the phonics group (M = 2.69, SD = 1.41) followed by the combined (M = 2.77, SD = 1.86) and whole word group (M = 2.83, SD = 1.19).

3.2.3. L1 Language-specific influence on L2 English

Our last research question aimed to investigate whether phonemic and orthographic differences in L1 German affect L2 English spelling accuracy and whether accuracy levels differ between groups after having received phonics, whole word or combined instruction for two years. Means, standard deviations and a one-way ANOVA with the target novel phoneme or orthographic pattern were calculated for two error categories across all three groups (Table 5). Error Category I consisted of four phonologically novel phonemes, /æ/, /θ/, /ð/ and /ɒ/, which do not exist in L1 German. Error Category II comprised words that are orthographically novel for L1 German speakers containing the silent <e> and /ʃ/ spelled as <sh>.

Errors were found in both categories stemming from phonological and orthographic differences between L1 German and L2 English, although the percentage of errors in each category differed. Analysing errors in Category I, learners had an average error percentage of 42% (M = 0.84, SD = 0.594) in sub-category I.1 (/æ/). Sub-category I.2 (/θ/) was where most errors were found across groups, with 92.67% of words spelled inaccurately (M = 1.85, SD = 0.356). Learners had an average error percentage of 67.33% (M = 1.35, SD = 0.797) in sub-category I.3 (/ð/), and with 18.00% (M = 0.36, SD = 0.561), fewest errors were made in sub-category I.4 (/ɒ/). Target words in Category II contained orthographically novel patterns, with sub-category II.2 (/ʃ/, 46.67% misspellings, M = 0.93, SD = 0.777) being more difficult to acquire for ELLs in comparison to sub-category II.1 (silent <e>, 27.33% misspellings, M = 0.55, SD = 0.759). Although learners in the phonics group outperformed their peers in the whole word and combined group on all six measures, only group differences for Category I.3 (/ð/) were statistically significant, F(2,72) = 5.23, p = .008. Learners in the phonics group spelled the words ‘this’ and ‘with’ more accurately compared to children in the two other groups, indicating less L1 influence from L1 German phonology. This suggests better mastery of L2 PGC.