Beginning to Read in Vietnamese: Kindergarten Precursors to First Grade Fluency and Reading Comprehension

Abstract

Measures of decoding and oral language have been shown to predict early reading comprehension across a wide variety of languages, though the timeframe and strength of the predictions vary by orthographic depth. This study is the first to examine predictors of early reading in Vietnamese, a transparent orthography of Romanized letters and diacritics. Eighty-two children in Hanoi, Vietnam, completed measures of decoding and oral language in kindergarten (phonological awareness, PA; rapid automatized naming, RAN; receptive and expressive vocabulary) and measures of decoding and reading comprehension in first grade. Average performance at the end of first grade, after just one year of formal instruction, was near ceiling on word reading but more variable on nonword and text reading. Kindergarten PA and RAN (but not vocabulary) predicted first-grade decoding after controlling for maternal education and kindergarten decoding, and PA was a stronger predictor than RAN (10% vs 4% of unique variance). The best predictors of first grade reading comprehension were first-grade decoding (47% of unique variance) and kindergarten expressive vocabulary (4% of unique variance) after controlling for maternal education. Overall, Vietnamese children became accurate and efficient decoders after one year of instruction. Findings from RAN and PA suggest their utility in guiding differentiated instruction on decoding. Kindergarten vocabulary, which differed as a function of maternal education, significantly predicted first-grade comprehension.

Children learning to read in different languages confront similar cognitive challenges, but the details of early reading, and thus the nature of early reading instruction, vary as a function of both the orthography used and features of the language represented. Most of the work on early reading development has focused on English, a language with a lexical stock drawn from several different sources and a relatively deep orthography. It typically takes English speakers two to three years to master all the decoding and spelling rules (e.g., Seymour et al., 2003), with considerable time in the early grades devoted to teaching decoding and spelling, especially irregularly spelled words (Ziegler & Goswami, 2005).

We know rather little about the course of early reading development in non-Indo-European languages, with the exception of Chinese (e.g., Anderson & Chen, 2013) and Korean (e.g., Kim, 2011) among Asian languages. Vietnamese, for example, a member of the Austroasiatic language family (Nguyen et al., 2018), uses the Roman alphabet with diacritics to represent its six tones and certain vowel singletons and combinations. The grapheme-phoneme mapping is relatively transparent, especially for speakers of the standard dialect (Pham & McLeod, 2016).

The study reported here is, to our knowledge, the first to look systematically at reading development in Vietnamese children. The present study was carried out as part of a larger study of early language and literacy development, for which 104 monolingual Vietnamese children in Hanoi, Vietnam, were tested on an array of language and emergent literacy skills in kindergarten before formal reading instruction had begun. Of these, 89 also completed assessments at the end of first grade. We report here on the kindergarten predictors of grade 1 reading, and the degree to which these predictive relationships, as well as concurrent relationships in grade 1, replicate findings from studies of children learning other alphabetic orthographies.

Reading Development and Predictors of Reading Outcomes Across Orthographies

We know from cross-national studies that word reading skills can develop very rapidly for readers of transparent orthographies given adequate instruction. For example, decoding instruction in transparent orthographies, including Finnish (Leppanen et al., 2006), Spanish (Goswami et al., 1998), Turkish (Öney & Durgunoğlu, 1997), and Kiswahili (Kim & Piper, 2019), often begins only when children are six years old (rather than in preschool or kindergarten as in the US) and is typically completed within one school year.

Another well-researched topic is how much the pattern of predictors of literacy outcomes attested in English varies as a function of orthographic transparency and/or different approaches to early literacy instruction. A meta-analysis conducted by the National Early Literacy Panel (NELP, 2008) found high continuity in conventional literacy skills over time for English speakers in the U.S. For example, decoding nonwords in kindergarten was the strongest predictor of decoding in first grade (Q = 0.72; NELP, 2008). Kindergarten precursors with moderate associations to English decoding include rapid automatized naming (RAN), phonological awareness (PA), oral language, and alphabetic knowledge (NELP, 2008). A cross-linguistic longitudinal study of English, Spanish, Slovak, and Czech (Caravolas et al., 2019) also found that word reading, RAN, and PA in kindergarten predicted decoding in first grade.

Nonetheless, the nature of precursors of decoding can vary across languages. For example, while PA is a robust predictor of decoding in English-speaking children (Bus & Van IJzendoorn, 1999; NELP, 2008), studies of children learning to read transparent alphabetic orthographies suggest that PA might be acquired as children learn to read, rather than being a prerequisite to reading (Landerl et al., 2019). While various precursors to decoding have been identified, vocabulary has been found to be a consistent predictor of reading comprehension across orthographies (e.g., Elleman, et al., 2009; Verhoeven & Perfetti, 2011). Vocabulary is also a good predictor of accurate reading of words in texts in deep orthographies like English, which often require resolving orthographic puzzles semantically (e.g., pronouncing dead, lead/lead, read/read correctly) (Elleman, et al., 2009).

The role of oral language as well as decoding in predicting reading comprehension was crystallized in the Simple View of Reading (SVR: Gough & Tunmer, 1986; Hoover & Gough, 1990), which recognizes the necessity of both components. Within the SVR, code-related skills are the predominant predictors of reading comprehension in grades 1 or 2, while decoding skills are still being built and while the texts read are linguistically simple; oral language skills emerge as important predictors in grades 3 or 4, as texts become more linguistically complex (Storch & Whitehurst, 2002). For English speakers, oral language and code-related skills are highly correlated in preschool or kindergarten, but then each contributes uniquely to reading comprehension starting in grade 3 (Language and Reading Research Consortium & Chiu, 2018).

For previously studied transparent orthographies, the shift from reliance on decoding to reliance on vocabulary in predicting comprehension occurs earlier. A meta-analysis comparing English to other Indo-European languages (Dutch, Greek, Spanish, French, Italian, Norwegian, and Finnish; Florit & Cain, 2011) found that oral language predicted reading comprehension much earlier for children learning transparent alphabetic orthographies than for children learning deep orthographies, who remain busy with learning to decode much longer.

Measures of Reading Accuracy and Fluency Across Languages

In tests of the SVR, the constructs of decoding and oral language have been measured differently across studies as well as across languages. In a meta-analysis of English reading (Garcia & Cain, 2014), the relation between decoding and reading comprehension was strongest when decoding was measured as untimed word reading accuracy. Nonetheless, all measures of decoding (word-level accuracy and speed) were highly predictive of reading comprehension.

For transparent orthographies, with near one-to-one letter-sound correspondence, decoding accuracy may play a less important role in predicting comprehension because most children achieve ceiling quite early. For such languages, decoding fluency may be more sensitive to developmental changes than accuracy alone (Caravolas et al., 2019). Decoding fluency reflects the automaticity of processing that is necessary to free up the cognitive capacity required to integrate the meaning of the larger text units (Fuchs et al., 2001; Silverman, et al., 2013). Florit and Cain (2011) found that, for transparent orthographies, the predictive relationship between decoding and reading comprehension was stronger when decoding was measured as (non)word fluency than as accuracy, and the relationship between fluency and reading comprehension was even more robust in younger readers who had fewer years of instruction. Decoding skills have been measured using (non)word fluency in the transparent orthographies of Africa including Kiswahili, Kikamba, and Lubukusu (Kim & Piper, 2019) as well as in European languages.

Fluency can be measured as speed of reading single (non)words as well as speed in reading passages. Passage reading or text reading fluency, defined by appropriate prosody as well as speed, makes contextual clues available and reflects attention to the syntactic units that must be parsed to achieve the meaning. Thus, oral language has been associated with text reading fluency in several languages including English (e.g., Kim et al., 2011), Korean (e.g., Kim, 2011), and three sub-Saharan African languages (Kim & Piper, 2019). Text reading fluency has been proposed as a measure of overall reading competence (Fuchs et al., 2001) and is widely applied in standardized tests of reading comprehension. Finally, the relation of word reading fluency to text reading remains an open question. Text reading either completely or only partially mediates the relation between word reading and reading comprehension in different studies (see Kim & Piper, 2019).

Reading Instruction in Vietnam

The present investigation focuses on Vietnamese, a new language in the scientific study of reading. Given the paucity of studies of instructional effects in Vietnam, we provide some background here on schooling and literacy instruction to contextualize the present study. Vietnam achieved universal access to primary education in 2002, by which time tests for 3^rd and 5^th graders were also administered in order to track progress and identify geographic areas that needed more resources (Griffin & Thanh, 2006). In 2014, Rolleston and Krutikova reported that Vietnamese students on average performed very well in comparison to other similar-income countries, but that large rural-urban disparities persisted, with urban students in the Red River Delta performing best.

Kindergarten programs are part of preschool education in Vietnam, and attendance is not compulsory (Vietnam National Assembly, 2005). Reading instruction begins in first grade, the first year of compulsory schooling, when children are six years old. The first grade reading program in Vietnam offers a systematic sequential introduction to the orthographic system. This program uses a single series of textbooks titled Tiếng Việt [Vietnamese] to teach language arts in all primary schools throughout Vietnam. The two textbooks used in first grade, Tiếng Việt 1: Tập 1 (Vietnamese 1: Book 1; Dang, Hoang, & Tran, 2014) and Tiếng Việt 1: Tập 2 (Vietnamese 1: Book 2: Dang, Hoang, Hoang, & Tran, 2014), consist of 103 rime-focused decoding lessons, (i.e., “Học vần” [study rimes]), followed by 13 weekly lessons to synthesize skills (i.e., “Luyện tập tổng hợp” [integrated practice]). The decoding lessons introduce pairs of singletons (vowel and consonant), followed by pairs of rimes until all rimes of the language have been reviewed. Each lesson includes exercises in identifying, reading, and writing rimes in isolation, with onsets to make single words, and in short paragraphs. The integrative lessons focus on one theme per week (e.g., school, family) and include activities of reading, writing, spelling, and story retell. Figure 1 displays a passage from Vietnamese 1: Book 2 that children are expected to read fluently by the end of first grade.

Figure 1. Reading Passage from First Grade Textbook.

Note. This reading passage is from Tiếng Việt 1, Tập 2 [Vietnamese 1, Part 2], page 142, taught in the second half of first grade. The following is a summary in English: An old man is in his garden planting a custard apple tree. His neighbour suggests that he plant a banana tree instead because he will be able to enjoy the fruit, whereas the custard apple tree takes longer to bear fruit. The old man responds that if he will not be able to eat the custard apples, then at least his grandchildren will. They will not be able to forget the person who planted the tree.

According to the benchmarks outlined by the Ministry of Education and Training in Vietnam (MOET, 2018), children by the end of first grade are expected to recognize all letters and rimes and to read aloud at a rate of 40-50 words per minute with appropriate phrasing, to answer basic comprehension questions, summarize the main idea and details of a story, and identify key features such as dialogue and characters’ actions. Notably, national decoding benchmarks in the early grades are defined using fluency (words per minute) rather than accuracy goals, in part because there is so little variation in accuracy scores.

Features of the Vietnamese Language and Orthography

Vietnamese has a transparent orthography with nearly one-to-one mapping of phonemes to graphemes. The phonotactic structure of Vietnamese words is quite limited, however; syllables can end only with one of three stops /p, t, k/, nasals /m, n, ŋ/, or vowels (Pham & McLeod, 2016). There are thus many words which are segmentally identical and differentiated only by tone (Pham et al., 2018). Additionally, the constrained phonotactics of Vietnamese can create a relatively complex visual array in which the diacritic details are crucial for differentiating words (as shown in Figure 1, a passage from the first-grade textbook). Furthermore, in many regional dialects of Vietnamese, multiple phonemes converge onto a single grapheme (for a review of dialects, see Pham & McLeod, 2016), which might create additional challenges for teaching children how to read and write. Though dialectical variation in pronunciation is important to consider, it is not directly addressed here, as this study was carried out in Hanoi, where the dialect is well aligned to the orthography.

Based on characteristics of Vietnamese orthography, we might expect that word reading accuracy can be achieved relatively easily, that fluency will be a better measure of decoding, and that oral language skills play an important role in predicting reading comprehension from an early point in reading development. Given the salience of tones in distinguishing Vietnamese lexical items, perhaps vocabulary predicts word reading more strongly and/or earlier in Vietnamese than in languages with more segmental variation. No formal instruction in the code is provided in Vietnamese kindergartens (Vietnam National Assembly, 2005), and the process of achieving automaticity in reading a language in which many words are distinguished only by diacritics might be delayed. Thus, we are particularly interested in determining how far children have progressed in decoding and reading comprehension at the end of one year of formal instruction. Following the guidance from previous studies of transparent orthographies (e.g., Florit & Cain, 2011; Kim & Piper, 2019), we measure reading fluency at both single item and text levels as measures of decoding.

The Present Study

This study examines the reading skills of monolingual Vietnamese children, in particular the kindergarten skills that predict first grade reading outcomes. To our knowledge, this is the first empirical study available on reading development in the Vietnamese language. The sample is from a larger longitudinal study (Pham et al., 2019), in which children living in Hanoi, Vietnam, completed oral language measures at the end of kindergarten and reading measures at the end of first grade.

One objective of the larger project was to develop a set of language and literacy assessment tools for Vietnamese that could be used by kindergarten teachers. Because reading instruction formally starts in first grade in Vietnam, children in kindergarten are not exposed to print instruction. Thus, we intentionally excluded literacy skills (e.g., letter name identification) in our kindergarten battery to avoid penalizing children with limited prior exposure to print. We also prioritized measures that can be quickly administered and scored with minimal training. Tasks of rapid automatized naming (RAN) and phonological awareness (PA) were prime candidates as predictors of decoding given their relative ease of administration and their well-established place in the cross-linguistic literature (e.g., Kirby et al., 2010; Landerl et al., 2019; Song et al., 2016). We also included measures of receptive and expressive vocabulary (e.g., Coll, 2005) in order to test the relevance of the SVR to Vietnamese. There were three research questions:

1. Following one year of literacy instruction, how do Vietnamese children perform on measures of decoding (indicated by reading fluency) and comprehension?

2. Which precursors in kindergarten predict fluency in first grade?

3. Which precursors in kindergarten predict reading comprehension in first grade?

Extrapolating from studies of other languages with relatively transparent orthographies (e.g., Seymour et al., 2003; Florit & Cain, 2011), we expect Vietnamese children to show accurate and fluent decoding by the end of first grade, following one year of reading instruction. Because fluency is highly correlated with reading comprehension (e.g., Fuchs et al., 2001), reading comprehension may also be high. Studies across alphabetic languages suggest that RAN is a consistent predictor of later reading whereas the role of PA can depend on the orthographic structure (Landerl et al, 2019). Oral language measures may play a smaller role than RAN or PA in predicting first-grade decoding (e.g., NELP, 2008).

Consistent with the Simple View of Reading (Gough & Tunmer, 1986), we anticipate that reading comprehension will be dependent on a combination of skills in decoding and oral language. Based on recommendations for transparent alphabetic orthographies (e.g., Florit & Cain, 2011; Landerl et al., 2019), this study measures decoding fluency (rather than accuracy). Fluency is measured at the single item level (i.e., nonword reading) and in a passage (i.e., text reading). While nonword reading and text reading are related but separable constructs (e.g., Kim, 2011), whether one (or both together) best predicts reading comprehension remains an open question (Kim & Piper, 2019). Of the oral language predictors, vocabulary is likely the most predictive of reading comprehension during this early stage (e.g., Elleman et al., 2009).

Method

Participants

As part of the larger project (Pham et al., 2019), 89 children completed measures at two time points. In order to focus on typical acquisition, the present study excluded 7 children due to a classification of developmental language disorder¹, leaving a total of 82 children (48 boys, 34 girls) with an average age of 5 years; 8 months at the start of the study. Children lived in Hanoi, Vietnam, were monolingual speakers of Vietnamese, and were from a wide range of SES, measured by maternal education (see Table 1).

Table 1.

Descriptive Statistics

	M	SD	Min	Max
Age (in months)	68.3	3.7	62.0	74.0
Maternal Education	4.0	1.0	1.0	6.0
Kindergarten Precursors
Receptive Vocabulary (max = 60)	53.5	2.8	43.0	58.0
Expressive Vocabulary (max = 60)	47.7	4.7	37.0	56.0
Phonological Awareness (max = 36)	12.1	6.0	0.0	34.0
RAN (correct items / second)	0.86	0.28	0.42	1.92
First Grade Reading
Nonword Fluency (max = 50)	35.6	13.5	5.0	50.0
Text Fluency (max = 190)	144.0	40.7	22.0	190.0
Reading Comprehension (max = 14)	10.0	2.6	0.5	13.0

Note. Maternal education is measured using a 6-point scale in which 0 = less than elementary and 6 = graduate school. Reading fluency was measured in words per minute. RAN = Rapid automatized naming of objects and digits.

General Procedures

Participants completed a set of measures in kindergarten and again at the end of first grade. In kindergarten, individual children completed tasks at their school sites in March or April 2017. In first grade, children completed tasks at a university center in Hanoi in June 2018, the first month of summer vacation. The study was approved by the Institutional Review Board at the first author’s university, and by kindergarten program principals and the university center in Hanoi. As part of the informed consent process, trained research team members provided information about the study in Vietnamese in written form and through verbal discussions over the phone and in person. Parents provided written consent and children provided written assent to participate.

Tasks were administered using PowerPoint slides on a computer laptop. For tasks that included audio stimuli (i.e., PA and receptive vocabulary), items were pre-recorded and presented simultaneously with the picture stimuli. Audio stimuli were presented to the child over headphones, with a research assistant also listening along with a separate set of headphones and a splitter.

Reading Tasks

Children completed tasks of reading fluency (word-level and text-level) and reading comprehension in first grade. Additionally, children completed tasks of reading fluency in kindergarten to verify reading status prior to the onset of formal instruction.

Word-Level Reading Fluency.

Word-level reading fluency was measured using tasks of word reading and nonword reading. Tasks were modelled after the Early Grade Reading Assessment (EGRA: Gove & Wetterberg, 2011), which has been adapted for a large number of languages including Vietnamese (Vu, Tran, & Tran, 2016). Children were shown a 5x10 array of items and were given one minute to read aloud as many as possible. Responses were digitally audio recorded for later scoring. Performance was measured as the number of items correctly read within one minute. Items that were read incorrectly or not read within the allotted time were considered errors.

To create the word recognition task, we started with the familiar word reading subtask of the Vietnamese EGRA (Vu et al., 2016). To confirm word familiarity, we calculated the frequency of each word using the Corpora of Vietnamese Texts (CVT: Pham et al., 2008), a language corpus of one million words collected from Vietnamese newspapers and children’s books. Word frequency was calculated as the natural logarithm + 1 of the raw count. We then excluded items that had values of 6 or less and then added high frequency words from the CVT (Pham et al., 2008). The final version of this task consisted of 50 test items with an average frequency of 8.5 and range of 7.5 to 10.1 (See Appendix A for a complete list). Items were presented in lower case letters and random order.

To create the nonwords, we started with two nonword recognition tasks from the Vietnamese EGRA (Vu et al., 2016), in which nonwords differed from real words by one phoneme (consonant, vowel, or tone). Using the CVT (Pham et al., 2008), we calculated word frequencies of all items and removed any item that had a frequency greater than zero (i.e., real words). Then, we eliminated words with low rime frequency to focus on rime structures commonly found in Vietnamese and filtered to avoid more than three items with the same onset. We eliminated duplicated nonwords and items that only differed by tone. We simplified the few visually complex nonwords (those with more than one diacritic marker) by reducing the number of diacritic markers and/or reducing the rime. When these items were simplified, 9 (of 50) became real words, which we retained because they were highly infrequent in the language (see Appendix B for a complete list). Items were presented in random order and in lower case letters.

Text Reading Fluency.

Two reading passages from the Vietnamese EGRA (Vu et al., 2016) were used to measure text reading fluency and reading comprehension. Children were asked to read two passages: Passage 1 consisted of 60 words and Passage 2 of 130 words. Children read passages aloud for a maximum of one minute each. Responses were digitally audio recorded to be scored at a later time. Because performance on Passages 1 and 2 was highly correlated (r = .62, p < .001), Passages 1 and 2 were summed to form the total score for text reading fluency, measured as the number of items correctly read within one minute (out of 190). See Appendices C and D for the two passages.

Reading Comprehension.

As per procedures outlined in the Vietnamese EGRA (Vu et al., 2016), children were given an opportunity to read silently Passages 1 and 2 after they had finished reading aloud. There was no time limit for silent reading. When ready, children answered comprehension questions about each passage while viewing the text to reduce the memory load (see Appendices C and D for the comprehension questions). Children’s responses were digitally audio recorded and transcribed using SALT software (Miller & Iglesias, 2012) for later verification of fluency scores and comprehension scoring. Reading comprehension was measured as the number of questions correctly answered for Passage 1 (out of 6) and Passage 2 (out of 8). Passage 1 and 2 were highly correlated (r = .54, p < .001), and thus were summed to form a total score for reading comprehension (out of 14). Cronbach’s alpha for the reading comprehension total was 0.79.

Kindergarten Precursors

Children in kindergarten completed PA, RAN, receptive vocabulary and expressive vocabulary. Performance on the vocabulary measures for the larger sample (N = 104) was previously reported in Pham et al., 2019.

Phonological awareness (PA).

The Vietnamese PA task created for this project consisted of four subsections: blending, tone detection, segmentation, and rime detection (See Appendix E for stimuli). In order to maintain child attention, children were presented with a PowerPoint presentation of a frog hopping on lily pads to get a prize (one prize per PA subsection). Each lily pad represented one item so that the child could see how many items were remaining to get the prizes. Children were presented with 1-2 practice items that could be repeated before starting each subsection, and on which children received feedback (correct or incorrect), but without any explanation. The dependent variable was the total number of correct items (maximum of 36). Cronbach’s alpha across all subsections was 0.84. The following is a brief description of each subsection.

Blending items asked children to combine an onset and rime to form syllables (4 items, e.g., s….ữa -> sữa [m…ilk -> milk]), two syllables to form a compound word (4 items, e.g., dưa…hấu -> dưa hấu [water…melon -> watermelon]), or three syllables to form a compound word (4 items, e.g., hình…tam…giác -> hình tam giác [tri…an…gle -> triangle]). Children completed 3 practice items (one before each blending type) and a total of 12 test items. Cronbach’s alpha for this subsection was 0.87.

Tone detection was modelled after a tone awareness task created for Chinese-speaking children (Siok & Fletcher, 2001) in which a child hears four syllables (e.g., to ti té ta), three of which contain the same tone, and were instructed to say the one that is different from the others (i.e., té). The placement of the target differed across items. Children completed 2 practice items and 6 test items. Cronbach’s alpha for this subsection was 0.75.

Segmentation consisted of dividing compound words into two (4 items) or three syllables (4 items) and producing a word without the initial phoneme (4 items). Before each test item, children were prompted to repeat the word before segmenting it to confirm they heard the target correctly (e.g., Nói “máy bay”. Bây giờ con nói “máy bay” mà không nói “bay” [Say “airplane”. Now say “airplane” without saying “plane”]). Children completed 3 practice items (one before each segmentation type) and a total of 12 test items. Cronbach’s alpha for this subsection was 0.86.

In rime detection, children listened to four words (e.g., banh, xanh, tanh, leo), three of which rhymed, and had to say the one that was different (i.e., leo). Children completed 2 practice items and 6 test items. Cronbach’s alpha for this subsection was 0.79.

Rapid automatized naming.

Rapid automatized naming (RAN) included naming objects and digits and was modelled after the rapid naming subtest of the Comprehensive Test of Phonological Processing (Wagner et al., 1999). For object naming, stimuli consisted of three animate objects (i.e., bear, cat, fish) and three inanimate objects (i.e., chair, house, star) to be sensitive to characteristics of Vietnamese classifiers (i.e., animacy = con; inanimacy = cái). Furthermore, we selected object names that did not vary across regional dialects. For digit naming, stimuli consisted of six single digits (i.e., 2, 3, 4, 5, 7, 8) presented in random order.

During practice, children first confirmed their familiarity with the six objects or digits by naming them aloud. Six kindergarten participants were not able to name digits during practice, and thus only completed object naming. Test items for objects and for digits each consisted of two arrays of 36 items each (4 rows of 9 pictures) in which the set of 6 items was presented in random order. Response time was measured from when the child began naming items until the end of the last item. The dependent variable was items per second, calculated as the total number of correctly named items (72 objects + 72 digits) divided by the total time in seconds.

Receptive Vocabulary.

A picture identification task was created for the larger project to measure receptive vocabulary. Children were shown a 2x2 array and asked to point to the picture that corresponded to a spoken word. Stimuli consisted of 60 objects, ranging from low frequency (to high frequency and presented in random order. The dependent variable was the number of items correctly named (maximum of 60). Cronbach’s alpha for the larger sample reported in Pham et al., 2019 was 0.71. See Appendix F for stimuli.

Expressive Vocabulary.

A picture naming task was created for the larger project to measure expressive vocabulary. Children were presented with a black-and-white line drawing on a computer screen and were asked to name the picture. Stimuli were 60 objects (nonoverlapping with the receptive vocabulary task) ranging from low to high frequency and presented in a random order. The dependent variable was the number of items correctly named (maximum of 60). Cronbach’s alpha previously reported in Pham et al., 2019 was 0.85. See Appendix G for stimuli.

Data Analysis

Because the sample was relatively small, we considered it important before undertaking the main analysis, to evaluate whether the data met statistical assumptions of linearity, independence of observations, homoscedasticity, and normal distribution. There was linearity as assessed by partial regression plots and a plot of studentized residuals against the predicted values. There was independence of residuals, as assessed by a Durbin-Watson statistic of 2.025, and no evidence of multicollinearity, as assessed by tolerance values greater than 0.1. There was homoscedasticity, as assessed by visual inspection of a plot of studentized residuals versus unstandardized predicted values. The assumption of normality was met, as assessed by Q-Q plots. Additionally, we evaluated potential outliers that had studentized deleted residuals greater than ±3 standard deviations. Because separate analyses with and without potential outliers resulted in the same results, we retained all data in the full analysis.

The main analysis consisted of descriptive statistics, correlations, and hierarchical multiple regression. We conducted descriptive statistics to evaluate mean performance and variability of each reading measure (word fluency, nonword fluency, text fluency, comprehension). We conducted bivariate correlations between the demographic variables (age, gender, maternal education) and reading measures to identify covariates for the regression models. Finally, we conducted separate hierarchical multiple regression to control for covariates and evaluate the relative contribution of each measure in predicting decoding and comprehension.

Results

Performance on Reading measures

As shown in Figure 2, performance on word and nonword fluency in kindergarten was at near floor levels. Following one year of reading instruction, average performance on word fluency in first grade was at ceiling levels (mean = 48 of 50 words/minute). Children had more difficulty in reading nonwords. As shown in Table 1 and Figure 2, mean performance on nonword fluency was 36 words/minute with wide variability (range from 5 to 50). Because of ceiling effects for word fluency, we focused on nonword fluency as the measure of word-level reading.

Figure 2. Reading fluency in kindergarten and first grade.

Note. Reading fluency displayed as the number of correct items per minute. Figure displays means and ±1 standard error bars.

For text-level fluency, children read Passage 1 fluently (mean = 57 of 60 words/minute). However, performance on Passage 2 was weaker (mean = 87 of 130 words/minute) and more variable than Passage 1 (SD = 35 vs. 8, respectively). As shown in Table 1, mean total scores were high (144 of 190 words/minute) though quite variable (range of 22 to 190 words/minute). Children answered an average of 5 of the 6 reading comprehension questions correctly on Passage 1 but only 5 of 8 questions correctly on Passage 2. Comprehension scores ranged from 0.5 to 13 across the two passages (maximum of 14; see Table 1).

Kindergarten Precursors of First-Grade Fluency

Table 2 displays a correlation matrix of all variables. Of the demographic variables collected (age, gender, maternal education), only maternal education was significantly related to the first grade reading measures. Moderate positive correlations between maternal education and reading measures ranged from r = .34 to .40. Of the kindergarten precursors, there was a strong correlation between expressive vocabulary and receptive vocabulary (r = .56, p < .001) and between RAN and PA (r = .50, p < .001). Notably, the vocabulary measures correlated with neither RAN nor with PA (see Table 2), suggesting that the language-related precursors measured separate constructs from the code-related precursors.

Table 2.

Bivariate Correlations between Variables

	1	2	3	4	5	6	7	8	9	10	11
1. Age	---
2. Gender	−.265*	---
3. Maternal education	−.114	−.178	---
4. Receptive vocabulary	.287**	−.111	.201	---
5. Expressive vocabulary	.271*	−.261*	.350**	.556**	---
6. Phonological awareness	.128	−.104	.312**	.169	.145	---
7. Rapid automatized naming	.253*	−0.085	.229*	0.114	0.141	.500**	---
8. Nonword fluency: K	.140	−.231*	.153	.021	.186	.608**	.399**	---
9. Nonword fluency: Gr1	.102	.015	.399**	.253*	.322**	.514**	.451**	.316**	---
10. Text fluency: Gr1	.112	.029	.361**	.365**	.410**	.565**	.519**	.349**	.886**	---
11. Read comprehension: Gr1	.087	−.015	.344**	.392**	.492**	.472**	.456**	.206	.629**	.771**	---

Note. Pearson correlations are displayed for precursors from kindergarten (K) and reading measures from K and first grade (Gr1). Gender was coded as 1= Female and 0= Male.

^*p < .05

^**p <.01

We conducted a hierarchical multiple regression model that controlled for maternal education and kindergarten nonword fluency and then entered all four kindergarten precursors in one block to compare their relative contribution to first grade nonword reading fluency. Then, based on the standardized Beta coefficients, each kindergarten precursor was entered into a hierarchical regression model from largest to least Beta value after controlling for the covariates of maternal education and kindergarten nonword fluency. As shown in Table 3, the full model accounted for 40% of the variance in first grade nonword fluency F(6, 75) = 8.343, p < .001. After controlling for maternal education and kindergarten reading (F(2, 79) = 11.62, p < .001, R² = .227), the two code-related precursors were significant: PA accounted for an additional 10% of unique variance (F(3, 78) = 12.71, p < .001, ΔR² = .101.), and RAN accounted for 4% of additional variance (F(4, 77) = 11.26, p < .001, ΔR² = .041). The vocabulary measures were not significant predictors of first grade word-level reading fluency.

Table 3.

Hierarchical Multiple Regression Models for Grade 1 Reading Fluency

	Nonword Fluency			Text Fluency
Variable	B	Beta	ΔR2	B	Beta	ΔR2
Phonological awareness: K	0.744*	0.332	0.101	2.489**	0.370	0.139
Rapid automatized naming: K	11.157*	0.229	0.041	42.275**	0.289	0.067
Expressive vocabulary: K	0.459	0.159	0.030	1.934*	0.223	0.076
Receptive vocabulary: K	0.229	0.048	0.001	1.879	0.131	0.011
Maternal education	2.595	0.184	0.160	3.449	0.081	0.130
Nonword fluency: K	−0.050	−0.034	0.067	−0.212	−0.049	0.088
R2	0.400			0.511

Note. Contributions of kindergarten (K) precursors to first grade reading fluency at word and text levels.

^*p < .05

^**p <.01.

A composite for text reading fluency was calculated by summing reading performance across Passages 1 and 2. As shown in Table 3, the full model accounted for 51% of the total variance in text reading fluency F(6, 75) = 13.059, p < .001. After controlling for maternal education and kindergarten reading (F(2, 79) = 11.003, p < .001, R² = .218), PA accounted for an additional 14% of unique variance (F(3, 78) = 14.433, p < .001, ΔR² = .139.); RAN accounted for 7% of additional variance (F(4, 77) = 14.14, p < .001, ΔR² = .067), and expressive vocabulary accounted for 8% of additional variance (F(5, 76) = 15.18, p < .001, ΔR² = .076). Receptive vocabulary was not a significant predictor of text reading fluency.

Kindergarten Precursors of First-Grade Reading Comprehension

The reading comprehension total score combined performance in answering comprehension questions from Passages 1 and 2. Because it is well documented that decoding plays a role in early reading comprehension (e.g., Gough & Tunmer, 1986), the first set of regression analyses addressed whether to include fluency of nonword reading, of text reading, or of both in the model to predict reading comprehension. A regression model of maternal education, kindergarten nonword reading fluency, and first-grade nonword reading fluency to predict reading comprehension was significant, R² = .40, F(3,78) = 17.171, p < .001: after controlling for maternal education and kindergarten reading, first-grade nonword reading fluency accounted for 26% of the total variance in reading comprehension. When first-grade text reading fluency was added, the model remained significant (F(4,77) = 30.159, p < .001, R² = .61), and text reading fluency accounted for 21% of additional variance.

In a separate series of regressions, a model of maternal education, kindergarten nonword reading fluency, and first-grade text reading fluency to predict reading comprehension was significant (R² = .59, F(3,78) = 37.860, p < .001): after controlling for maternal education and kindergarten reading, first-grade text reading fluency accounted for 45% of the total variance. When first grade nonword reading fluency was added to this model, it did not account for additional variance (p = .07). Finally, when comparing two models, (1) maternal education and first-grade text reading fluency and (2) maternal education, first-grade text reading fluency, and kindergarten nonword reading fluency; kindergarten nonword reading fluency did not account for unique variance beyond what was accounted for by first-grade text reading fluency alone (p = .35). Thus, maternal education and first-grade text reading fluency were the two control variables included in the next set of hierarchical multiple regressions to test the relative contribution of the kindergarten precursors.

As shown in Table 4, the full model consisted of maternal education and first grade text fluency as control variables, followed by the kindergarten precursors individually entered in the same order as the previous analysis (Table 3). This full model accounted for 63% of the total variance in reading comprehension, F(6,75) = 21.396, p < .001. After controlling for maternal education and first-grade text reading fluency (F(2, 79) = 56.45, p < .001, R² = .588), the only significant kindergarten precursor was expressive vocabulary, which accounted for 4% of unique variance beyond what was accounted for by maternal education and text reading fluency (F(5,76) = 25.85, p < .001, ΔR² = .037). PA and RAN measured in kindergarten were not significant predictors of first grade reading comprehension.

Table 4.

Multiple Regression Model for Grade 1 Reading Comprehension

Variable	B	Beta	ΔR2
Phonological awareness: K	0.027	0.061	0.002
Rapid automatized naming: K	0.775	0.082	0.003
Expressive vocabulary: K	0.110*	0.197	0.037
Receptive vocabulary: K	0.044	0.047	0.001
Maternal education	0.047	0.017	0.118
Grade 1 Text fluency	0.038**	0.582	0.471
R2	0.631

Note. Contributions of kindergarten (K) precursors to first grade reading comprehension.

^*p < .05

^**p <01.

Discussion

The purpose of this study was to evaluate reading performance by Vietnamese children following one year of formal instruction and to identify longitudinal predictors. These findings expand our understanding of early reading development by adding information about Vietnamese, a previously unstudied language/orthography, to the literature.

As is almost universally found, variability in reading competence is associated with maternal education level (the proxy for SES in this sample). Maternal education positively correlated with word-level reading, text reading, and reading comprehension (see Table 2). Even in a transparent orthography such as Vietnamese, maternal education can influence children’s early reading abilities. One way to account for the role of maternal education is through investigating home literacy practices. In a longitudinal study of Finnish, also a highly transparent orthography, Silinskas and colleagues (2020) found that more highly educated mothers were more likely to engage in shared reading activities, which in turn predicted children’s oral language skills and later reading fluency.

Kindergarten measures of word and nonword reading fluency confirmed that this sample of Vietnamese children had minimal literacy skills at the onset of the study (Figure 2). Floor performance in kindergarten provided verification that reading instruction had not yet begun. A few kindergarteners demonstrated some decoding skills, suggesting further study of precocious reading (for review, see Olson et al., 2006) as well as variation in home literacy practices (Silinskas et al., 2020) is needed for Vietnamese.

By the end of first grade, most Vietnamese children could decode familiar words relatively accurately and quickly (Figure 2), confirming findings from children learning other relatively transparent orthographies (Landerl et al., 2019) that word reading skill can be mastered quickly. Though these children were reading words accurately and fluently, discrepancy in performance between their word and nonword reading suggests that their decoding skills were still developing. The relatively small correlations between kindergarten vocabulary measures and first-grade nonword reading (e.g., r = .32 between nonword fluency and expressive vocabulary) as well as the lack of predictive relationships between the kindergarten vocabulary measures and first-grade nonword reading fluency (see Table 3) reflect the fact that children cannot rely on their vocabulary knowledge or word familiarity to read nonwords.

Consistent with previous findings across languages (e.g., Caravolas et al., 2019; Landerl et al., 2019), RAN and PA in kindergarten predicted first-grade decoding fluency. However, the strength of each precursor to decoding has varied across studies. Landerl et al. (2019) posited that PA was a strong predictor only for opaque orthographies. However, our results on Vietnamese, a transparent orthography, also found that PA was a stronger predictor than RAN (see Beta values in Table 3). The positive correlation between PA and kindergarten decoding (Table 2) in the absence of explicit instruction, combined with the predictive power of kindergarten PA on first-grade decoding fluency (Table 3), suggests that Vietnamese-speaking children may acquire PA skills in the process of learning to read. If so, their development replicates that of the speakers of five transparent Indo-European languages in which PA was not a precursor (i.e., unidirectional relation from PA to reading) but rather a consequence of reading (i.e., unidirectional relation from reading to PA) (Landerl et al., 2019).

Expressive vocabulary in kindergarten was a significant predictor of first grade text reading fluency but not of nonword reading fluency. This finding confirms that text reading involves the integration of decoding and background knowledge (Fuchs et al., 2001), and identifies text reading fluency as itself a proxy for reading comprehension (as suggested by Francis et al., 2018) rather than being a pure decoding measure.

Interestingly, we found that text reading fluency accounted for all of the decoding variance in reading comprehension. Word-level reading, as measured by nonword fluency, did not account for additional unique variance beyond that accounted for by text reading fluency. The relative importance of word-level fluency versus text fluency on reading comprehension can depend on the target language. Studies across languages have found either complete mediation in which text fluency fully mediates the relation of word-level reading to reading comprehension (e.g., two of three African languages studied in Kim & Piper, 2019) or partial mediation in which word-level fluency is directly related to reading comprehension as well as indirectly related via text fluency (e.g., third African language in Kim & Piper, 2019; see Kim, 2011, for similar results in Korean). Our results are consistent with complete mediation models. Future studies are needed to replicate this finding in a larger sample of Vietnamese children and to examine the role of text-level reading fluency across orthographies of varying depth.

Consistent with the Simple View of Reading (Gough & Tunmer, 1986), decoding skills were highly related to reading comprehension in the beginning stages of reading. Perhaps most striking was the shift in the predictive strength of the kindergarten precursors within this single age-group of children: whereas RAN and PA were significant predictors of first grade text fluency (Table 3), oral language (expressive vocabulary in particular) was the only significant kindergarten predictor of first grade reading comprehension (Table 4). This shift has been reported as well in English readers, but over a greater time span (e.g., Storch & Whitehurst, 2002). The earlier shift found in transparent Indo-European languages (Florit & Cain, 2011; Caravolas et al., 2019) is replicated here for Vietnamese.

Study Limitations and Future Directions

There are a few study limitations to acknowledge. First, the measure of word fluency reached ceiling performance. As one of the first measures of word fluency developed for the Vietnamese language, we targeted high frequency words to be administered to first graders (who have had only one year of reading instruction). Future development of tasks to measure word accuracy and fluency could include more items and a wider range of frequency (from high to low).

This study identified overall PA as a strong predictor of reading. To strike a balance between task length and maintaining child attention, each PA skill was measured using a limited number of items. Future studies with more items to measure each PA skill are needed to examine which specific skills within PA are most sensitive to predicting reading outcomes. We also recognize that reading comprehension in the early school years relies heavily on extracting concrete information that is readily available in the text. Future studies of reading comprehension in Vietnamese, particularly for older ages, should include questions that require higher level comprehension skills such as synthesis and inference.

Finally, the degree of orthographic depth can also vary by regional dialect, as certain dialects are more directly aligned with the writing system than others (Pham & McLeod, 2016). The relatively quick acquisition and accurate decoding found in this sample may not be replicated in additional samples, particularly among children who speak different regional dialects of Vietnamese. Studies of how dialectal differences influence reading efficiency can enhance the literature on early reading development beyond language typology alone.

Educational Implications

This study on early reading in Vietnamese contributes to the knowledge base on transparent orthographies beyond Indo-European languages. Vietnamese is an interesting test case for reading development as the orthography has a near one-to-one sound-letter correspondence though can become visually complex with the addition of tone and vowel diacritic markers.

Findings of this study also have implications for school curriculum and for future investigation of atypical populations. Although the first-grade curriculum includes decoding of onsets and rimes, phonological awareness as a skill itself is not explicitly included in the national standards and benchmarks in Vietnam that shape the curriculum. Evidence from other transparent orthographies suggests that children develop PA as a product of learning to read rather than needing it as a precursor (Landerl et al., 2019). Nonetheless, the inclusion of PA benchmarks and related classroom activities in Vietnam may be particularly useful for some children.

These results make clear, though, that the major source of individual differences in reading comprehension will be more closely related to students’ vocabulary knowledge than to their decoding, a skill on which the majority can be expected to reach ceiling, especially in light of the well-structured and systematic teaching materials that are used in Vietnamese schools. Teaching vocabulary, and the content and concepts that vocabulary skills index, should receive substantial educational attention in the early grade classrooms (e.g., Elleman et al., 2009) particularly to reduce the differences associated with maternal education documented here.

Finally, by the end of first grade, Vietnamese children are expected to decode all onset and rime combinations, even syllables that are low frequency or have many diacritic markers (e.g., MOET, 2018). Children who are not able to decode single words fluently by the end of first grade may need intervention or differentiated instruction until they meet age norms; in such a transparent orthography it should be possible to identify children in need of early reading intervention very early in their trajectories toward skilled reading, thus preventing cumulative reading difficulties. RAN and PA may be effective quick screening tools to identify children at risk for later reading difficulties, and further research that includes samples of children with reading difficulties are needed to extend findings presented here.

Appendix

Appendix A.

Stimuli for Word Recognition Task

Item	Raw	Ln + 1	English
1. và	23695	10.07	and
2. có	18907	9.85	yes
3. đã	14009	9.55	already
4. của	11418	9.34	of
5. một	8110	9.00	one
6. đến	8021	8.99	to
7. các	7866	8.97	the
8. như	7521	8.93	as
9. ông	7423	8.91	man
10. không	7383	8.91	no
11. trong	7244	8.89	in
12. ở	7219	8.88	at
13. cho	7159	8.88	for
14. nhà	6657	8.80	house
15. làm	6544	8.79	do
16. người	6487	8.78	people
17. được	6395	8.76	can
18. con	6284	8.75	child
19. tôi	5926	8.69	me
20. sẽ	5818	8.67	will
21. ta	5596	8.63	we
22. nhiều	5486	8.61	many
23. những	5451	8.60	these
24. nam	5413	8.60	south
25. với	5296	8.57	with
26. lên	4911	8.50	up
27. cả	4359	8.38	both
28. để	4083	8.31	to
29. vào	3762	8.23	enter
30. này	3758	8.23	this
31. năm	3711	8.22	year
32. ra	3635	8.20	out
33. về	3615	8.19	about
34. họ	3591	8.19	them
35. nước	3589	8.19	water
36. trường	3576	8.18	school
37. khi	3540	8.17	when
38. mình	3502	8.16	body
39. tại	3248	8.09	in
40. nhất	3173	8.06	best
41. phải	3141	8.05	right
42. chỉ	2939	7.99	just
43.thấy	2932	7.98	see
44. rồi	2416	7.79	done
45. cái	2282	7.73	the
46. cô	2162	7.68	miss
47. nếu	2147	7.67	if
48. tới	1972	7.59	next
49. giờ	1944	7.57	hour
50. ăn	1885	7.54	eat

Note. Items are listed from most to least frequent. Raw counts were calculated from the Corpora of Vietnamese Texts (Pham et al., 2008) which consists of over one million Vietnamese words collected from newspapers and children’s books. The English gloss serves as a reference point only as certain Vietnamese word classes (i.e., classifiers) cannot be directly translated into English.

Appendix B.

Stimuli for Nonword Recognition Task

Item	Raw	Ln + 1
1. ná	5	1.79
2. thạ	0	0
3. mêm	0	0
4. òn	2	1.10
5. ăm	4	1.61
6. nhạ	0	0
7. thẻm	0	0
8. đỏi	0	0
9. rĩn	0	0
10. điệm	0	0
11. xọa	0	0
12. hon	31	3.47
13. giàn	60	4.11
14. gáng	0	0
15. choa	2	1.10
16. quẽ	0	0
17. mun	1	0.69
18. hêng	0	0
19. phừ	0	0
20. chảu	0	0
21. dụn	0	0
22. đẹ	0	0
23. hản	0	0
24. ém	18	2.94
25. giõ	0	0
26. găn	0	0
27. tón	0	0
28. tợ	0	0
29. phậm	0	0
30. hánh	0	0
31. hãn	18	2.94
32. ọa	0	0
33. hiệt	0	0
34. nỏi	0	0
35. hìn	0	0
36. gó	0	0
37. lêm	0	0
38. quạy	0	0
39. đươn	0	0
40. ngãy	0	0
41. bơn	0	0
42. sèo	0	0
43. bũng	0	0
44. hị	0	0
45. hía	0	0
46. xứi	0	0
47. ện	0	0
48. khẹ	0	0
49. trủn	0	0
50. khưng	0	0

Note. Frequency counts were calculated using the Corpora of Vietnamese Texts (Pham et al., 2008). All but 9 items had a frequency of zero, demonstrating that items were not real words. The 9 items with raw counts greater than zero were obscure (low frequency).

Appendix C. Reading Passage 1

Bà cho bé Mai một con gà rất đẹp.

Mỗi sáng, Mai đều chờ gà gáy.

Nhưng nó không gáy.

Một buổi trưa, con gà kêu: “Cục ta cục tác.”

Mẹ nói với Mai: “Con gà mái của con mới đẻ trứng đấy!”

Bây giờ bé Mai đã biết vì sao con gà này lại không gáy ò ó o.

1. Ai cho Mai con gà?

2. Mỗi sáng, ai chờ con gà gáy?

3. Con gà của Mai kêu thế nào?

4. Con gà của Mai kêu cục ta cục tác vào buổi nào?

5. Mẹ nói gì với Mai?

6. Vì sao con gà của Mai không gáy?

English translation

Grandma gave Mai a very pretty chicken.

Every morning, Mai waited for it to crow.

But it did not crow.

One afternoon, the chicken said “buck buck buck”.

Mom told Mai: “Your hen just laid an egg!”

Now, Mai knows why this chicken does not crow.

1. Who gave Mai the chicken?

2. Every morning, who waited for the chicken to crow?

3. How did Mai’s chicken sound?

4. When did Mai’s chicken say buck buck buck?

5. What did mom say to Mai?

6. Why didn’t Mai’s chicken crow?

Note. Passage is from the Vietnamese version (Vu et al, 2016) of the Early Grade Reading Assessment (EGRA: Gove & Wetterberg, 2011) used to measure text reading fluency and reading comprehension.

Appendix D. Reading Passage 2

Sang và con mèo xám

Tên tôi là Sang, vừa tròn tám tuổi. Em trai

tôi là Châu, bốn tuổi. Chúng tôi thích chơi đùa

với con mèo nhà mình. Con mèo màu xám,

to xù và thích trốn sau tủ.

Một hôm, con mèo biến đâu mất. Chúng tôi

nghĩ nó đang chơi trốn tìm,

nhưng không thấy nó ở những chỗ thường chơi.

Vì thế, chúng tôi lục khắp nhà để tìm con mèo .

Cuối cùng, chúng tôi thấy nó ở gầm giường,

nhưng nó không chỉ có một mình! Nó đẻ ba con

mèo con: hai con màu xám , một con màu trắng.

Khi chúng tôi kể về những con mèo con,

bố mẹ nói mẹ cũng sẽ sinh em.

Chúng tôi sắp có một em gái!

1. Ai đang kể câu chuyện này?

2. Vì sao con mèo to xù?

3. Em trai của Sang tên là gì?

4. Sang và Châu thích làm gì?

5. Tại sao Sang và Châu nói con mèo không ở đó một mình?

6. Theo con tại sao con mèo biến đi mất một lúc?

7. Con mèo đẻ mấy con?

8. Sang và Châu tìm thấy con mèo ở đâu?

English translation

Sang and the grey cat

My name is Sang, I am eight years old. My younger brother

is Chau, four years old. We like to play

with our house cat. The cat is gray,

big and likes hiding behind shelves.

One day, the cat disappeared. We

thought it was playing hide and seek,

but we did not see it in the places it usually plays.

So we searched everywhere in the house for a cat.

Finally, we found it under the bed,

But it is not alone! It gave birth to three

kittens: two were grey, one was white.

When we told our parents about the kittens,

they said that mother will also be giving birth.

We are about to have a little sister!

1. Who is telling this story?

2. Why is the cat big?

3. What is name of Sang's younger brother?

4. What do Sang and Chau like to do?

5. Why did Sang and Chau say the cat was not there alone?

6. Why do you think the cat disappeared for a while?

7. How many kittens did the cat have?

8. Where did Sang and Chau find the cat?

Note. Passage is from the Vietnamese version (Vu et al, 2016) of the Early Grade Reading Assessment (EGRA: Gove & Wetterberg, 2011) used to measure text reading fluency and reading comprehension.

Appendix E.

Phonological Awareness Measure in Vietnamese (PhAM-V)

Section and Instructions	Target response
Ghép Tiếng [Blending] Con sẽ nghe các phần của những từ riêng biệt. Con hãy nối các phần thành từ. [You will hear parts of a word. Connect the parts to say the word.]
Practice: cà…chua	cà chua
1. bàn …ghế	bàn ghế
2. bác …sĩ	bác sĩ
3. áo… dài	áo dài
4. dưa….hấu	dưa hấu
Practice: kính…thiên…văn	kính thiên văn
5. xanh…da…trời	xanh da trời
6. câu…lạc…bộ	câu lạc bộ
7. bánh…mì…thịt	bánh mì thịt
8. hình… tam… giác	hình tam giác
Practice: b…ướm	bướm
9. h…ọc	học
10. s…ữa	sữa
11. g…ạo	gạo
12. đ…ọc	đọc
Nhận thức thanh điệu [Tone detection] Bây giờ con sẽ nghe bốn từ. Con hãy nói từ nào khác hẳn so với các từ còn lại. [Now you will hear four words. Say the word that is different from the others.]
Practice: to , ti , té , ta	té
Practice: mở, mủ, mả, mẹ	mẹ
1. nô, nỉ , no, na	nỉ
2. kẹ, kị, kí , kệ	kí
3. hẹ, hệ, họ, hà	hà
4. rỉ, rỏ, ra , rổ	ra
5. vù , vo, vi, ve	vù
6. xo, xu, xa, xè	xè
Phân chia [Segmentation] Nói …. Bây giờ nói … mà không nói …. [Say …. Now say … without ….]
Practice: xe đạp…xe	đạp
Practice: máy bay …bay	máy
1. gia đình…đình	gia
2. quần áo…áo	quần
3. trường học…học	trường
4. nhà cửa…nhà	cửa
Practice: hướng dẫn viên…viên	hướng dẫn
5. kính thiên văn…kính	thiên văn
6. mũ bảo hiểm…hiểm	mũ bảo
7. câu lạc bộ…lạc	câu bộ
8. câu trả lời…trả	câu lời
Practice: tìm…t	ìm
9. chăn…ch	ăn
10. thông…th	ông
11. nhân…nh	ân
12. bút…b	út
Nhận thức về vần [Rime detection] Bây giờ con sẽ nghe bốn từ. Con hãy nói từ nào khác hẳn so với các từ còn lại. [Now you will here four words. Say the word that is different from the others.]
Practice: banh, xanh, tanh, leo	leo
Practice: hâm, loa, xâm, câm	loa
1. náo, triết, miết, viết	náo
2. sinh, minh, tôi, đinh	tôi
3. pháp, táo, ngáp, nháp	táo
4. mỗi, bữa, sữa, nữa	mỗi
5. ghe, lông, trông, không	ghe
6. hoà, loà, vần, xoà	vần

Appendix F.

Receptive Vocabulary

	Item	English translation	Raw frequency	Ln + 1
Practice A	Tiền	Money	1196	7.09
Practice B	Đường	Road	1144	7.04
1	Bọ cạp	Scorpion	1	0.69
2	Bút màu	Crayons	2	1.10
3	Kính hiển vi	Microscope	2	1.10
4	Chữ nhật	Rectangle	2	1.10
5	Cằm	Chin	4	1.61
6	Nha sĩ	Dentist	4	1.61
7	Bong bóng	Balloon	5	1.79
8	Tã	Diaper	5	1.79
9	Chanh	Lemon / Lime	7	2.08
10	Ngựa vằn	Zebra	8	2.20
11	Tôm hùm	Lobster	13	2.64
12	Con nhím	Porcuppine	14	2.71
13	Chìa khoá	Keys	15	2.77
14	Con cua	Crab	17	2.89
15	Tê giác	Rhinocerous	19	3.00
16	Cá voi	Whale	19	3.00
17	Nấm	Mushroom	25	3.26
18	Nông trại	Farm	25	3.26
19	Phong bì	Envelope	29	3.40
20	Sách	Book	29	3.40
21	Mì	Noodles	30	3.43
22	Lâu đài	Castle	30	3.43
23	Vòi	Hose	31	3.47
24	Gối	Pillow	32	3.50
25	Giếng	Well	34	3.56
26	Chuối	Banana	35	3.58
27	Dây chuyền	Necklace	37	3.64
28	Lá cờ	Flag	38	3.66
29	Con sói	Wolf	40	3.71
30	Con nai	Deer	42	3.76
31	Quà	Gift/ Present	61	4.13
32	Bếp	Stove	69	4.25
33	Cá sấu	Alligator	70	4.26
34	Mây	Cloud	74	4.32
35	Đậu	Beans	74	4.32
36	Tô	Bowl	88	4.49
37	Tóc	Hair	97	4.58
38	Con ong	Bee	105	4.66
39	Cái thang	Ladder	106	4.67
40	Xương	Bone	136	4.92
41	Cơm	Rice	157	5.06
42	Trống	Drum	167	5.12
43	Mưa	Rain	180	5.20
44	Quần	Pants	198	5.29
45	Lỗ tai	Ear	270	5.60
46	Giấy	Paper	272	5.61
47	Khăn	Towel	273	5.61
48	Con trai	Boy	292	5.68
49	Con gà	Chicken	293	5.68
50	Lá	Leaf	294	5.69
51	Con mèo	Cat	306	5.73
52	Bác sĩ	Doctor	339	5.83
53	Con chuột	Mouse	360	5.89
54	Áo	Shirt	369	5.91
55	Con chim	Bird	407	6.01
56	Con gấu	Bear	474	6.16
57	Con thỏ	Rabbit	480	6.18
58	Vua	King	549	6.31
59	Bàn	Table	549	6.31
60	Con cá	Fish	632	6.45

Note. Items ranged from low to high frequency and were administered in random order.

Appendix G.

Expressive Vocabulary Stimuli

	Item	English translation	Raw Frequency	Ln+1
Practice A	Con chó	Dog	272	5.61
Practice B	Ghế	Chair	212	5.36
1	Xe lửa/ tàu lửa/ tàu điện	Train	1	0.69
2	Con công	Peacock	1	0.69
3	La bàn / bàn ủi	Iron	2	1.10
4	Xe cứu thương / xe cấp cứu	Ambulance	2	1.10
5	Con cá ngựa	Seahorse	2	1.10
6	Cái ô/ cây dù	Umbrella	2	1.10
7	Chim cánh cụt	Penguin	4	1.61
8	Máy chụp hình / máy chụp ảnh	Camera	5	1.79
9	Lược	Comb	5	1.79
	Sở thú / Vườn thú/ Vườn bách
10	thú	Zoo	5	1.79
11	Mắt kính / mắt kiếng	Glasses	6	1.95
12	Trạm xăng /cây xăng	Gas station	6	1.95
13	Kim cương	Diamond	6	1.95
14	Lính cứu hoả / Người cứu hoả	Fireman	9	2.30
15	Nến/ đèn cầy	Candle	10	2.40
16	(Con) ốc sên	Snail	10	2.40
17	(Con) chuột túi	Kangaroo	12	2.56
18	(Con) bọ rùa	Ladybug	14	2.71
19	(Máy bay) trực thăng	Helicopter	16	2.83
20	(Con) thiên nga	Swan	18	2.94
21	Quạt	Fan	22	3.14
22	Yếm	Bib	24	3.22
23	(Con) ruồi	Fly	28	3.37
24	Kẹo	Candy	28	3.37
25	(Con) nhện	Spider	34	3.56
26	Tim	Heart	35	3.58
27	Râu	Beard	38	3.66
28	Cửa sổ	Window	48	3.89
29	(Con) lạc đà	Camel	49	3.91
30	(Quả) dâu / (trái) dâu	Strawberry	52	3.97
31	(Quả) táo/ (trái) táo/ (trái) bôm	Apple	64	4.17
32	Đinh/ốc	Nail / Screw	68	4.23
33	(Con) đại bàng	Eagle	76	4.34
34	Em bé	Baby	78	4.37
35	Não/óc	Brain	81	4.41
36	Điã/ dĩa	Plate	88	4.49
37	Đồng hồ	Watch	90	4.51
38	Ngô/ Bắp	Corn	103	4.64
39	Giường	Bed	110	4.71
40	(Con) lợn/ (con) heo	Pig	151	5.02
41	(Con) ngựa	Horse	153	5.04
42	(Con) dao	Knife	153	5.04
43	Ly/cốc	Glass / Cup	173	5.16
44	(Con) cú	Owl	213	5.37
45	Máy bay	Airplane	254	5.54
46	Núi	Mountain	272	5.61
47	(Con) rắn/ (con) tranh	Snake	281	5.64
48	Bánh (ngọt)	Cake	292	5.68
49	(Con) hổ/ (con) cọp	Tiger	364	5.90
50	Thuyền/ tàu	Ship / boat	622	6.43
51	Kéo	Scissors	701	6.55
52	Cây	Tree	961	6.87
53	(Con) kiến	Ant	973	6.88
54	Tất/ vớ	Sock	1366	7.22
55	Xe/ xe hơi/ xe ô tô/ ô tô	Car	1539	7.34
56	Tay	Hand	1732	7.46
57	Hoa/ bông	Flower	2045	7.62
58	Cầu	Bridge	2165	7.68
59	Hình/ tranh	Picture	2204	7.70
60	Nhà	House	6645	8.80

Note. Items ranged from low to high frequency and were administered in random order. Parentheses indicate words that can be included in the child’s response (e.g., classifiers) but are not required for full credit. Back slashes separate words that can vary by regional dialect and are considered correct.