Journal of Experimental Child Psychology 126 (2014) 245–263

Contents lists available at ScienceDirect

Journal of Experimental Child Psychology journal homepage: www.elsevier.com/locate/jecp

The moderating effect of orthographic consistency on oral vocabulary learning in monolingual and bilingual children Kathleen Jubenville, Monique Sénéchal ⇑, Melissa Malette Department of Psychology, Carleton University, Ottawa, Ontario K1S 0W4, Canada

a r t i c l e

i n f o

Article history: Received 10 July 2013 Revised 3 May 2014

Keywords: Vocabulary Reading Orthographic consistency Children Learning Literacy acquisition

a b s t r a c t Two studies were conducted to assess whether (a) the incidental presence of print facilitates the acquisition of oral vocabulary, (b) the facilitative effect of print is moderated by phoneme-to-grapheme consistency, and (c) the findings obtained with monolingual children generalize to bilingual children. In total, 71 monolingual French-speaking children (Mage = 9 years 2 months) in Study 1 and 64 bilingual children (Mage = 9 years 3 months) in Study 2 participated in one of three conditions: consistent print, inconsistent print, or no print. Children were to learn novel labels for unfamiliar objects in a paired-associate paradigm. In both studies, print facilitated the acquisition and recall of expressive vocabulary. The effect of print consistency, however, varied across studies. As expected, monolingual children exposed to consistent print learned more novel labels than children exposed to inconsistent print. In contrast, bilingual children exposed to inconsistent print learned and recalled more labels than children exposed to consistent print. These intriguing findings might be due to differences in attention allocation during training. Ó 2014 Elsevier Inc. All rights reserved.

Introduction Vocabulary knowledge has been positively and robustly linked to the amount of reading elementary school children do (e.g., Sénéchal, 2006), and yet we are only beginning to investigate how ⇑ Corresponding author. E-mail address: [email protected] (M. Sénéchal). http://dx.doi.org/10.1016/j.jecp.2014.05.002 0022-0965/Ó 2014 Elsevier Inc. All rights reserved.

246

K. Jubenville et al. / Journal of Experimental Child Psychology 126 (2014) 245–263

reading might support oral vocabulary acquisition. Central to the current research was the notion that written representations can facilitate oral word learning for those who can read. Considering the complementary nature of spoken and written language, orthography may provide an additional source of analogous information to support oral word learning. For children and adults, hearing novel words in context can result in the creation of phonological representations in memory that are linked to semantic referents. Similarly, reading novel words in context also provides opportunities to acquire orthographic representations and link those representations to semantic referents (Share, 1999). When phonological and orthographic representations co-occur, they may become bonded in the mental lexicon and, therefore, increase the quality of word form representations as well as the availability of their contents (Ehri, 1978; Perfetti & Hart, 2002). Of primary importance for learning, orthographic information might facilitate the encoding of more precise phonological information for literate individuals (Rosenthal & Ehri, 2008). The facilitative effect of print on learning vocabulary might be especially valuable for bilingual learners, whose vocabulary tends to lag behind that of monolingual learners (Bialystok, 2008). The current research was conducted to test whether written information, presented incidentally, would facilitate oral vocabulary learning and retention in literate monolingual and bilingual children. Although an orthographic facilitation effect on oral vocabulary acquisition has been documented with first- and second-language adult learners (e.g., Zarei, 2009), there are few studies testing the effect in children. An early set of studies by Ehri and Wilce (1979) demonstrated that Grade 1 and Grade 2 children learned more spoken labels when paired with spelling than when paired with symbols or single letters. Recently, Rosenthal and Ehri (2008) examined whether an orthographic facilitation effect would be present when print was presented incidentally. In separate studies, Grade 2 and Grade 5 children who could read were taught paired associations between spoken words and their meanings. The children heard spoken rare nouns paired with simplified oral definitions that drew on familiar knowledge as well as picture representations. The presence of print during training was manipulated; half of the pictured items included the word in print and half did not. Importantly, no attention was drawn to the print. Results from both studies were similar in that the incidental presence of print enhanced children’s acquisition of expressive vocabulary across trials and increased recall of pronunciation and spelling accuracy 1 day later. Moreover, the facilitative effect of print on oral vocabulary learning has been replicated with samples of Chinese Grade 3 children learning English as a second language (Hu, 2008), typically developing children and literate children with Down syndrome (Mengoni, Nash, & Hulme, 2013), and literate children with autistic spectrum disorder (Lucas & Norbury, 2013). In three studies, the researchers also showed that the print facilitation effect was due to the print itself rather than a cueing effect (i.e., a condition in which the print was substituted with abstract symbols) (Ehri & Wilce, 1979; Hu, 2008; Mengoni et al., 2013). In sum, there is a growing body of evidence showing that print provides a congruent and complementary support to oral vocabulary learning. The words used in the above studies, with the exception of the science words in Lucas and Norbury (2013), were consistent and regular. English, however, is an opaque orthography characterized by multiple mappings between phonemes and graphemes (Katz & Frost, 1992). The less transparent the mapping, the more difficult it might be for literate individuals to exploit correspondences between print and speech. Therefore, it is possible that the orthographic facilitation of oral word learning might be influenced by the degree to which speech-to-spelling mappings are consistent. We found one published study that examined this issue. In addition to manipulating the presence of orthography during oral word learning, Ricketts, Bishop, and Nation (2009) manipulated phoneme-to-grapheme consistency (i.e., whether phonemes map onto one or more graphemes). Children were asked to learn three types of novel spoken labels: labels with consistent spelling, labels with inconsistent vowel spelling, and labels with inconsistent consonant spelling. Although Ricketts and colleagues’ findings seemed to show that inconsistent spelling affected oral vocabulary during the learning phase, the researchers did not find a statistically significant orthography-by-consistency interaction for production trials or for recognition reaction times at posttest. Hence, no strong conclusion could be made. It is possible, however, that the lack of a clear orthography-by-consistency effect might be due to the research design used. The design included only two items in each of the orthography-present conditions and only three learning trials that resulted in children producing less than 50% of items at the end of training. Moreover, posttesting occurred immediately following the learning session and was limited to

K. Jubenville et al. / Journal of Experimental Child Psychology 126 (2014) 245–263

247

recognition memory. Therefore, the current research was conducted to reexamine the effect of orthographic consistency by increasing the number of items per condition, increasing the number of learning trials, including a 1-day delay between learning and posttesting, and measuring recall at posttest. The current research was conducted in French. Both French and English have orthographic systems that are considered to be opaque; however, the source of their opacity differs. Whereas most inconsistencies in English are due to vowels, most inconsistencies in French are due to silent consonants at the end of words (Perry, Ziegler, & Zorzi, 2014). To illustrate, a comparison was conducted using French (BRULEX: Content, Mousty, & Radeau, 1990) and English (CELEX: Baayen, Piepenbrock, & Gulikers, 1995) lexical databases. When comparing monosyllabic words with inconsistent consonant–vowel–consonant (CVC) endings, silent letters were the source of phoneme-to-grapheme inconsistencies in 94% of French cases as compared with less than 1% of English cases. In the current research, the role of phoneme-to-grapheme consistency on oral word learning was examined by comparing children’s performance with words that did or did not include silent consonant endings. It is noteworthy that silent letter endings frequently carry morphological information in French. In fact, when derivational and inflectional morphological markers are not taken into account, speech-tospelling inconsistencies usually disappear (Peereman, Sprenger-Charolles, & Messaoud-Galusi, 2013). This makes the study of French interesting because it is one of the only languages in which the written language carries more morphological information than the spoken language. In addition, the predominantly silent morphology means that children must learn how to spell these parts of spoken words correctly without an oral reference, increasing the difficulty of the task (Sénéchal, Basque, & Leclaire, 2006). In the current research, Grade 3 children learned novel labels for pictures of novel objects (as in Ricketts et al., 2009). As in previous research, it was hypothesized that literate children who were exposed incidentally to orthography (both consistent-print and inconsistent-print conditions) would learn and retain more expressive vocabulary as compared with children in a no-print condition (i.e., a print facilitation effect). It was also hypothesized that transparent phoneme–grapheme mappings would facilitate connections between print and speech; therefore, children in the consistent-print condition were expected to learn and recall more novel labels than children in the inconsistent-print condition (i.e., a print consistency effect). In addition, children were trained and tested on receptive vocabulary, and their performance was expected to be at ceiling as in Ricketts and colleagues’ (2009) study. At posttest, children were also asked to spell the novel labels to assess whether children were acquiring orthographic forms (Ricketts et al., 2009; Rosenthal & Ehri, 2008, 2011). Spelling provided a verification of the print consistency manipulation with the expectation that consistent words would be easier to spell correctly than inconsistent words (Sénéchal et al., 2006). Two studies were conducted in the current research. Study 1 was conducted with monolingual French-speaking children, and Study 2 was conducted with bilingual children. Children in the two studies were in the same classrooms in French schools. Studying whether findings obtained with monolingual speakers generalize to bilingual children is important for practical and theoretical reasons. In their classrooms, teachers use the same techniques to instruct monolingual and bilingual (or multilingual) children. Yet, bilingual children lag behind their monolingual peers in word knowledge (Bialystok, 2008). As reviewed in Study 2, there is theoretically driven research showing that bilingualism has cognitive advantages and disadvantages that might alter the predicted print facilitation and print consistency effects.

Study 1 Method Participants Monolingual French-speaking children were recruited from seven Grade 3 classrooms in three schools where the language of instruction was French. The schools were located in a large city and an adjacent town in the province of Québec, Canada. Québec is the only Canadian province where French is the majority language. Of the 72 monolingual French-speaking children recruited, 71 had normal hearing and one had a hearing problem according to parent reports. Only children with normal

248

K. Jubenville et al. / Journal of Experimental Child Psychology 126 (2014) 245–263

hearing were included in the analyses. Children (56% girls and 44% boys) were, on average, 9 years 2 months of age (SD = 2 months). Parents reported their education level and household income. The majority of respondents reported having completed a university degree, with 44% having completed an undergraduate degree and 34% having completed a postgraduate degree. This level of educational attainment was higher than that of the city–town and province from which the sample was drawn (Statistics Canada, 2007). The distribution for the sample (city–town and province for individuals aged between 25 and 64 years) was as follows: 1% with some high school (vs. 5%–16% city–town and 17% province), 3% with high school diploma (vs. 15%–21% city–town and 21% province), 18% with college diploma or certificate (vs. 26%–32% city–town and 36% province), and 78% with university degree (vs. 25%– 47% city–town and 21% province). Of the 54 parents who reported their family income, the distribution was as follows: 82% with household income above $90,000, 9% between $70,001 and $90,000, 6% between $50,001 and $70,000, 3% between $30,001 and $50,000, and 1% between $0 and $30,000. The income for these families was generally higher than the provincial median for couples with children ($76,339) but was representative of the city–town ($91,692–$134,371) from which it was drawn (Statistics Canada, 2007). Design Children were matched in triplets on their total reading score (± three words) and assigned randomly within triplets to one of three learning conditions. In the consistent-print condition, novel labels with consistent phoneme-to-grapheme correspondences were printed below the pictured referent. In the inconsistent-print condition, novel labels with inconsistent phoneme-to-grapheme correspondences (due to ending with a silent consonant) were printed below the pictured referent. In the no-print condition, there were no printed labels. The novel labels were phonologically identical in the consistent-print and no-print conditions. The main hypotheses were tested with a priori orthogonal contrasts; the test of a print facilitation effect compared the two print conditions with the no-print condition, and the test of a print consistency effect compared the consistent-print condition with the inconsistent-print condition. Stimuli To control for prior lexical knowledge, the novel labels were nonwords. Two lists of six two-syllable decodable nonwords with legal French spellings were constructed: one with more consistent spelling and the other with more inconsistent spelling. Orthographic consistency (speech-to-spelling correspondence) was manipulated by either including or not including a silent consonant at the end of the nonwords. Importantly, the second syllable was pronounced identically in all conditions but was spelled differently in the consistent-print and inconsistent-print conditions. The use of two-syllable nonwords was necessary to equate items across conditions on the number of letters and phonemes. All items had six letters and five phonemes. The consistent nonwords had a CVV/CCV letter structure with a vowel digraph in the first syllable (e.g., poncra /pO~ka/), and the inconsistent nonwords had a CV/CCVC structure with a silent consonant ending (e.g., pocrat /poka/). To make the items within a list as distinct as possible, the six two-syllable nonwords in each list began with a unique consonant. Consonant blends were also unique to each list. Each of the three possible vowel pairings for the first syllable (i.e., o /o/ with on /O~/, u /y/ with ou /u/, e /E/ with eu /ø/) and each vowel in the second syllable (i.e., a, i, o) was employed only twice within a list. The resulting nonword pairs, their pronunciations, and the number of phonological and orthographic neighbors are listed in the Appendix. To construct the nonwords, common French word part units were selected from a database of French words (LEXOP: Peereman & Content, 1999) according to their monosyllabic–bidirectional consistency properties and frequency characteristics. Importantly, all nonwords with orthographies that existed in English or French were eliminated, as were those that could be perceived as sounding like real words. In addition, to help equate memory demands across items, nonword constructions were limited to those in which no letter or sound occurred more than once and to pairs of nonwords that had a similar number of phonological and orthographic neighbors (Roux & Bonin, 2013). Independent samples t tests revealed no significant differences across print conditions in the mean number of

K. Jubenville et al. / Journal of Experimental Child Psychology 126 (2014) 245–263

249

phonological neighbors, orthographic neighbors, or unit frequencies of final syllables (ps > .50). As desired, mean grapheme-to-phoneme consistency of final syllables, based on token count, was not statistically significantly different across print conditions, t(5) = 1.46, p = .20, whereas phonemeto-grapheme consistency was significantly different, t(5) = 3.10, p = .03, with a higher mean level of consistency favoring the consistent-print condition. To verify that nonwords did not resemble existing French words, and to authenticate that they looked and sounded as though they could be French words, nonwords were piloted with adult French speakers. Ten French first-language speakers were asked to spell to dictation 11 pairs of matched consistent–inconsistent items. The adults were then provided with a list of nonwords and asked to indicate whether (a) the spellings could exist in French, (b) the pronunciations could exist in French, and (c) they resembled any real French words. Based on the results, six consistent–inconsistent word pairs were chosen to use in the current studies. In addition, a French first-language adult confirmed, on hearing them, that final syllables across word pairs were identical. Picture referents. Three-dimensional colored pictures of unfamiliar or novel objects, used in Ricketts and colleagues’ (2009) study, served as fixed referents for the nonwords to be learned. The same pictures were used across all three learning conditions: six for items to be learned and a seventh as a training item. Word definitions were not included because they could introduce potential confounds such as differences in working memory demands across nonword definitions (length/complexity) and the unintentional lexical priming of other words in children’s memory that may facilitate or interfere with the association to be learned. Learning paradigm A paired associate word-learning paradigm was adapted from Ricketts and colleagues (2009) and Rosenthal and Ehri (2008). The learning session consisted of a minimum of six and a maximum of nine training cycles; each cycle included a repetition block followed by a production block. Each block consisted of six trials (i.e., one trial per nonword). In each block, the six nonwords were presented in a different preset randomized order. The learning criterion was three accurate successive blocks of production trials. The number of training cycles and criterion for success were consistent with those used by Rosenthal and Ehri with Grade 2 children and was chosen in light of Ricketts and colleagues having reported low recall after three training cycles. During learning, children received feedback after an unsuccessful repetition or production trial, and this feedback consisted of a condition-specific repetition trial. This is in contrast to Ricketts and colleagues and Rosenthal and Ehri, where feedback was given after each trial irrespective of accuracy. Central to this work, the two print conditions had the target labels printed below each pictured referent. The nonword was written in lowercase letters in black Arial font (size 40). Although the print was presented incidentally in these conditions, children who read the word had an additional exposure to labels. To prevent this as a potential confound, labels in the no-print condition were provided twice orally in training and feedback trials (Rosenthal & Ehri, 2008). Audio-recordings and pictures were presented on a laptop or an iPad in order to ensure consistency of presentation and timing across items and testers. The recordings were those of a French first-language male adult using a neutral tone of voice and a reasonable articulation speed. Pilot testing with two children confirmed that they could repeat each nonword after hearing its recorded label. Training procedure. Children participated individually. At the beginning of training, the experimenter showed a picture of an alien and explained that the children would learn some made-up words and that the words stood for things that an alien might use. The children then completed a block of repetition trials. They were told that they would see a picture on the screen and hear the name of the object in the picture spoken. They were asked to listen carefully because they would be asked to repeat it. The experimenter then initiated the practice trial with the practice item valcron (/valkO~/). Following a 5-s exposure, the picture disappeared automatically and children were prompted to provide the spoken label. If children did not provide an accurate response or provided a response prior to the prompt, additional instructions were given by the experimenter to ensure understanding.

250

K. Jubenville et al. / Journal of Experimental Child Psychology 126 (2014) 245–263

Once children acknowledged their understanding of the instructions, the experimenter proceeded with the first cycle of repetition and production blocks. If children were correct on a repetition trial, they were provided with encouragement before proceeding to the following item. If the children’s response was inaccurate, they received encouragement and the trial was repeated once. Following the completion of a repetition block, a black screen appeared and the experimenter proceeded to the production block with these instructions: ‘‘Now, I’ll show you all of the pictures again, one at a time. The pictures will be mixed up, so look carefully. I want you to try and tell me the name of the object in each picture. If you’re not sure, give me your best guess.’’ All responses received encouragement, whereas inaccurate responses received feedback that consisted of a repetition trial for that item. Following the completion of the production block, a black screen appeared before beginning the next training cycle. Measures Reading measures. It was important to balance reading skill across conditions because Rosenthal and Ehri (2008) found that it interacted with children’s ability to learn in the print conditions. Regular and irregular word reading was assessed with subtests of the Batterie d’Évaluation du Langage Écrit et de ses Troubles (Mousty, Leybaert, Alegria, Content, & Morais, 1994). Each subtest required that children read aloud 24 isolated words matched on frequency and length. Because the word cassis in the irregular word reading test had two correct pronunciations, it was excluded; consequently, irregular word reading was scored on 23 words. Learning. Children’s learning consisted of their performance on the repetition and production blocks during each of the nine learning cycles. Expressive vocabulary posttest. Children’s recall of nonwords was assessed by asking children to produce the spoken label for each picture. The pictures, in grayscale on 6  6-inch cards and without any orthography, were presented one at a time to children. Changing the nature of the stimulus by presenting it on a card instead of a computer screen, and in grayscale instead of color, provided a means of assessing whether children were able to transfer their acquired knowledge beyond the learning environment. Cards were presented one at a time in a randomized order. Children were asked to name the pictured object, and if they did not respond within 5 s the experimenter provided encouragement and proceeded to the next card. Children received encouragement but no feedback. Receptive vocabulary posttest. To assess whether children recognized the spoken labels of each picture referent, they were asked to select, from a 2  2 array of four pictures, the picture of the object being named. The task was presented on a laptop or an iPad, and the same audio-recordings as during training were used. The experimenter recorded children’s selections. Because the arrays were constructed from the six pictures used during learning, it was important to counterbalance each array to minimize guessing correctly. In addition, it decreased the potential for children’s decisions on each array to interfere with their subsequent identification on arrays that followed (Ricketts et al., 2009). To do so, a random sequence generator (Haahr, 1998) was employed to determine possible unique combinations of four integers, and from these six possible target combinations were created with target areas counterbalanced across combinations. Arrays were presented in a fixed randomized order, but the target sequence was determined systematically such that an equal number of children across conditions completed each of the possible combinations. Spelling dictation posttest. To assess children’s implicit acquisition of word-specific orthographies, children were asked to spell to dictation each nonword learned the previous day. Children heard the recording of each nonword twice. Nonwords were presented in a randomized order for each child, and spelling was scored for accuracy and types of errors. Procedure Recruitment and testing took place at the end of two consecutive school years, that is, during May and June. Children participated individually in a quiet area within their school. Testing was completed

251

K. Jubenville et al. / Journal of Experimental Child Psychology 126 (2014) 245–263

over three sessions held on consecutive days. On Day 1 children completed reading measures, on Day 2 they completed the learning phase, and on Day 3 they were posttested. Posttests were administered in the following fixed order: expressive vocabulary (a recall task), receptive vocabulary (a recognition task), and spelling dictation. Responses were scored online, and sessions were audio-recorded to allow for verification and correction of scoring. Data verification Online scoring of children’s performance was compared with the audio-recordings to verify accuracy. These checks revealed that during training experimenters had mistakenly indicated that 12 children (17%) had reached criterion (i.e., three accurate and consecutive production cycles) and stopped the training when in fact these children had mispronounced at least one label during one of the production blocks. For these children, the mode for reaching criterion was the seventh cycle out of a maximum of nine cycles. The consequence of this experimenter error is that these children did not receive feedback on their error trials, and this may have attenuated their performance. However, 10 of these 12 children were distributed equally across the two print conditions, that is, in the conditions hypothesized to be most conducive to learning. Therefore, including their corrected data in the statistical analyses would not bias results in favor of the print facilitation hypothesis. That is, an attenuated performance would work against a print facilitation effect. In addition, full data verification could not be performed for eight children because some of their responses were inaudible, and for one additional child all responses were missing. In these instances, online scoring was used. Although these experimental errors did not bias results, they may have added unnecessary noise to the dataset; therefore, two dummy variables were created to indicate these cases and were used in the analyses. Findings with these covariates are reported. Results and discussion As shown in Table 1, children across conditions were reading at similar levels (ps > .45). Matching children on reading resulted in similar mean ages across conditions (p > .18) as well as parent education and family income (ps > .63). Therefore, any print effect found could not be due to these variables. Other preliminary analyses modified subsequent analyses in three ways. First, examination of children’s reading performance revealed that two children had irregular word reading performance that was more than 2 standard deviations below the sample mean. Including these children’s performance in the analyses changed the pattern of findings; consequently, their performance was excluded. Second, we found that irregular word reading interacted with learning condition for expressive

Table 1 Numbers of participants, mean ages, and reading performance (and standard deviations) as a function of study and learning condition. Study and variable

Condition Consistent print

Inconsistent print

No print

M

(SD)

M

(SD)

M

(SD)

Study 1 Number Age in years Regular word reading Irregular word reading

23 9.1 23.2 17.0

(0.5) (1.1) (4.6)

23 9.1 22.7 17.2

(0.4) (1.9) (3.4)

23 9.3 22.9 16.9

(0.4) (1.2) (4.6)

Study 2 Number Age in years Regular word reading Irregular word reading

20 9.3 23.0 16.6

(0.4) (1.2) (3.2)

20 9.3 22.7 15.9

(0.3) (1.3) (3.7)

22 9.2 22.9 16.3

(0.5) (1.7) (3.4)

Note. There were 24 and 23 words in the regular word reading and irregular word reading tasks, respectively.

252

K. Jubenville et al. / Journal of Experimental Child Psychology 126 (2014) 245–263

vocabulary learning; consequently, it was used as a covariate in the analyses (ps > .03). Third, only 20% of children reached criterion and typically did so on the seventh production block. Of these children, 10 were in the two print conditions and four were in the no-print condition. Given the low numbers, no further criterion analyses were conducted.

ax = 6

Receptive vocabulary During each learning cycle, children were asked to repeat each spoken label after hearing it while viewing its picture cue. As shown in the top panel of Fig. 1, children in the two print conditions (i.e., consistent print and inconsistent print) had no difficulty in repeating novel labels, with their performance near or at ceiling. In contrast, children in the no-print condition had more difficulty. Given the ceiling effect on the two print conditions, no statistical test of print consistency was conducted. Fig. 1 also presents the posttest receptive vocabulary performance (1-day delay). As is clearly shown in the graph, performance here was near or at ceiling in all conditions. This suggests that, after sufficient exposure, children had no difficulty in selecting the correct image for each label from an array of four

ax = 6

Posest recognion

Posest recall

Learning cycle: Producon

Fig. 1. Mean numbers of nonwords as a function of learning condition in Study 1. Top panel: Nonwords repeated during training and recognized at posttest. Bottom panel: Nonwords produced during training and recalled at posttest. Error bars represent standard errors.

K. Jubenville et al. / Journal of Experimental Child Psychology 126 (2014) 245–263

253

pictures that included the pictures for three other labels. Having established that children in each condition had learned an association between the oral labels with their pictured referents, we examined whether print would facilitate the formation of precise phonological representations that could be recalled. Expressive vocabulary After each repetition block during the learning phase, children were shown each picture cue (without print) and asked to label them one at a time. The bottom panel of Fig. 1 shows the accuracy of children’s production of spoken labels during this learning phase. These data were analyzed with a mixed design analysis of covariance (ANCOVA) with condition (consistent print, inconsistent print, or no print) as a between-participant variable and learning cycle (nine cycles) as a within-participant variable. The covariates were irregular reading performance and the two dummy variables identifying experimental errors. This analysis revealed a significant main effect of condition, F(2, 63) = 24.80, MSE = 7.04, p < .001, g2 = .44. The print facilitation contrast was significant, showing that children in the two print conditions produced more spoken labels accurately during learning than children in the no-print condition, contrast estimate = 1.54, confidence interval (CI) = [1.08, 1.99], p < .001, Cohen’s d = 1.55. The print consistency contrast was also statistically significant, contrast estimate = 0.54, CI = [0.01, 1.60], p < .05, Cohen’s d = 0.98. As expected, the incidental presence of consistent print facilitated production more than the presence of inconsistent print. The analyses also revealed a Cycle  Condition interaction, F(16, 504) = 5.61, MSE = 0.78, p = .001, g2 = .15. Follow-up polynomial contrasts were assessed for each condition separately (Bonferroni-corrected p < .02). These analyses yielded significant linear trends in all conditions (ps < .001), and this reflected children’s gradual learning. However, there were also significant quadratic trends in the two print conditions (ps < .001) because the sharpest increases were in the initial trials and the smallest increases were in the final trials. No other contrast was significant. Fig. 1 also presents posttest recall performance. A one-way ANCOVA, with the same covariates as the previous analysis, yielded a main effect of condition, F(2, 63) = 14.80, MSE = 2.21, p < .001, g2 = .32. The print facilitation contrast was significant, showing that children in the two print conditions retained more labels than children in the no-print condition, contrast estimate = 2.10, CI = [1.33, 2.87], p < .001, Cohen’s d = 1.43. However, the effect of print consistency was no longer statistically significant on posttest (p = .44). Hence, print consistency affected learning but did not have a lasting effect on recall. Spelling In the current research, children’s spelling to dictation was collected to assess whether print consistency would affect the acquisition of accurate orthographic representations. Nonwords across conditions were intentionally designed to have matching final syllables; therefore, children’s spelling accuracy on the first and second syllables was analyzed separately to show that the locus of difficulty was the addition of a silent consonant ending on this second syllable. A multivariate analysis of variance (MANOVA) for posttest spelling of the first and second syllables yielded a main effect of condition, Wilks’ k(4, 130) = .161, p < .001. Children in both the consistentprint condition (M = 5.9, SD = 0.3) and inconsistent-print condition (M = 5.8, SD = 0.4) had no difficulty in spelling the first syllable of the novel labels correctly (p = .62). In contrast, and as expected, differences across conditions appeared on the second syllable; children in the consistent-print condition spelled the second syllable accurately (M = 5.7, SD = 0.7), whereas children in the inconsistent-print condition had more difficulty with the syllable that contained the silent letter (M = 2.6, SD = 1.5), contrast estimate = 3.17, CI = [2.34, 4.00], p = .001, Cohen’s d = 2.67. An error analysis revealed that 95% of second-syllable errors made by children in the inconsistent-print condition were silent letter errors, with 51% of children omitting the silent consonant altogether and 49% substituting it with another letter. This suggests that letters that have no phonological value are more difficult to encode orthographically. Despite the fact that children in the no-print condition never saw the written labels, their spelling performance was included in the above analyses. These children spelled more than half of the labels as

254

K. Jubenville et al. / Journal of Experimental Child Psychology 126 (2014) 245–263

they were written in the consistent-print condition (Ms = 3.6 and 3.7, SDs = 1.2 and 1.8, for the first and second syllables, respectively). The performance of children in the no-print condition is consistent with children’s oral performance. That is, children must have accurate phonological representations of words in order to capture novel words in writing. In sum, the analyses of spelling errors in the inconsistent-print and no-print conditions highlight the interplay between phonology and orthography. Supplemental analyses First, we examined the theoretical claim that the presence of print facilitates the construction of accurate phonological representations for literate individuals. To do so, we assessed the stability of children’s spoken label production errors during learning and on subsequent posttests. A stable erroneous representation was defined as making the same phonological error on the last three production cycles of the learning phase (e.g., saying noudri instead of loudri). These error analyses revealed that there were more stable erroneous representations in the no-print condition (21%) than in the consistent-print condition (4%). There were no stable erroneous representations in the inconsistent-print condition, and the remaining production errors in all conditions were unstable (i.e., not maintained during the last three production cycles). We then examined whether children who had stable erroneous representations would carry over the representations on the delayed recall and spelling posttests. In the no-print condition, most children maintained their erroneous representations on recall (80%, 12/15) and spelling (87%, 13/15). In contrast, only four children across the print conditions had stable erroneous representations; of these children, three maintained their erroneous representations at recall, whereas one recalled and spelled them correctly. These analyses provide additional evidence that orthography can help to clarify phonological representations in memory. The second supplemental analysis examined the relation between word production during training and recall on posttest by assessing whether the probability of recalling a word was affected by the number of times children produced it during learning (i.e., production trials). This item analysis revealed a steplike increase in probability of recall for words produced from one to eight times during training, and these probabilities were .20, .37, .39, .53, .84, .88, .92, and .94, respectively. In contrast, labels that were not produced during training had only a .12 probability of being recalled. As has been shown with younger children (Sénéchal, 1997), learning expressive vocabulary seems particularly sensitive to successful retrieval practice. Study 2 The study of bilingualism is of interest for practical and theoretical reasons. Practical reasons are straightforward; more than half of the world’s population speaks more than one language (United Nations Educational, Scientific, and Cultural Organisation [UNESCO], 2003), 21% of children schooled in the United States speak more than one language (Aud et al., 2011), and more than 30% of the children in the region where this study was conducted speak more than one language (Statistics Canada, 2012). Hence, understanding how bilingualism might affect learning is warranted given that teachers are likely to have both monolingual and multilingual children in their classrooms. An increased understanding is particularly important in light of the established finding that bilingual children and adults show a vocabulary and lexical retrieval weakness as compared with their monolingual counterparts (e.g., Bialystok, 2008; Bialystok, Craik, & Luk, 2008). There are three key theoretical issues in the study of bilingualism that are relevant to the current research. One issue is whether there are cognitive consequences of bilingualism. The work of Bialystok and colleagues has shown that bilinguals seem to have more cognitive control than their monolingual counterparts (Bialystok et al., 2008). Another central issue is whether lexical knowledge (semantic, phonological, and orthographic) in both languages is automatically activated when using a single language. Here, studies with young adults have shown that performance can be slowed down due to this automatic activation of orthographic representations in two languages (for a detailed description of the parameters of this effect, see Jared & Kroll, 2001). Crucial to our research is evidence that orthographic knowledge develops quickly in the languages spoken. For instance, Grade 2 children learning to read in French and English can discriminate whether spelling patterns of nonwords (e.g., sotre vs.

K. Jubenville et al. / Journal of Experimental Child Psychology 126 (2014) 245–263

255

crump) are more likely to occur in French or English (Jared, Cormier, Levy, & Wade-Woolley, 2013). Another issue has been whether some linguistic or literacy components are language general or language specific. Studies with school-aged children have shown that skills such as phoneme awareness and alphabet knowledge seem to be language general because performance in one language is an excellent predictor of literacy skills in the other language (e.g., Jared, Cormier, Levy, & WadeWoolley, 2011; Laurent & Martinot, 2010). The evidence on the language specificity of orthographic knowledge is mixed and seems to depend on the degree of overlap between languages. In research on English–French bilinguals, orthographic processing in both languages was a unique predictor of reading in each language (Deacon, Wade-Woolley, & Kirby, 2009). In sum, there is evidence that bilingualism affects language and literacy performance. This body of research suggests that the findings obtained with monolingual children in Study 1 might not necessarily generalize to bilingual children and that at least three possible differences might occur. First, the lexical deficits exhibited by bilingual children might make word learning more difficult, and this would result in a lower performance in all conditions as compared with the monolingual children in Study 1. Second, and alternatively, if bilingual children have more cognitive control, then they might be able to take better advantage of the print, resulting in a larger print effect than was found in Study 1. Third, if spelling–sound correspondences for the languages spoken are automatically activated, then there will be a conflict between the silent consonant endings that occur in French but not in other languages. That is, the pronunciation of the novel items in the inconsistent-print condition maps onto the French orthography only. The resolution of this conflict would require additional processing and attentional resources that might modify the inconsistent-print effect found in Study 1. On the one hand, if children do not have enough time to resolve the conflict, then we might see a larger inconsistent-print effect than in Study 1. On the other hand, if children have enough time to resolve the conflict, then the increased processing resources might have enhanced the orthographic and phonological representations for the novel items with inconsistent print, thereby reducing or reversing the inconsistent-print effect. Study 2 was conducted as an initial step in understanding how bilingualism might affect the print facilitation and print inconsistency effects observed in Study 1. Method Participants In total, 64 bilingual and multilingual Grade 3 children (47% girls and 53% boys) participated. Children were recruited at the same time and from the same schools as those of Study 1. The children were, on average, 9 years 3 months of age (SD = 3 months). Parents indicated the languages spoken at home, with 83% of families being bilingual (i.e., 64% French and English and 19% French and another language) and the remaining 17% families being multilingual (i.e., 12% with three languages at home that included French and 6% with two other languages at home and the children learning French at school). Importantly, the bilingual and multilingual children did not differ in regular and irregular word reading (ps > .87). The languages spoken were all written with an alphabetic script, mostly a Latin script and, in four families, an Arabic script. Of note, three families spoke a tonal language spoken in South Eastern Africa that is sometimes transcribed with a defective Latin script because the script does not capture the tonal feature of the language. Finally, the written systems of the other spoken languages do not include silent consonants at the end of words (in fact, only Portuguese has silent letters, but usually within words). As in Study 1, most parents had completed a university degree (79%), whereas 14% had completed college or vocational training, 5% had completed high school, and 2% had not finished high school. Of the 56 parents who reported their household income, 48% indicated that their income was over $90,000, 25% between $70,001 and $90,000, 9% between $50,001 and $70,000, and 18% between $0 and $30,000. Procedure and stimuli The procedures and stimuli of Study 1 were used. In elaborating the stimuli, special care had been taken to create nonwords that were more characteristic of French than of English. That is, there were no English words that sounded or looked like target nonwords.

256

K. Jubenville et al. / Journal of Experimental Child Psychology 126 (2014) 245–263

Data verification Online scoring of children’s performance was compared with the audio-recordings to verify accuracy. These checks revealed that during training experimenters had mistakenly indicated that 13% (8/64) had reached criterion (i.e., three accurate and consecutive production cycles) and stopped the training when in fact these children had mispronounced at least one stimulus during one of the final production blocks. For these children, the mode for reaching criterion was the seventh cycle out of a maximum of nine cycles. Most children (7/8) were in one of the two print conditions. As in Study 1, children’s corrected data (i.e., their score from the verification instead of the online false ceiling score) were used in the statistical analyses because doing so would not bias results in favor of the print facilitation hypothesis. Moreover, training verification could not be performed fully on an additional eight children because parts of the audiotapes were not audible. As a final check that these experimental errors did not alter the findings, dummy variables were created and entered as covariates in the statistical analyses. These covariates did not alter the pattern of statistical significance; consequently, the analyses are reported without these covariates.

Results and discussion Preliminary analyses revealed that a criterion analysis was not necessary because only 8% of children reached the learning criterion and typically did so on the eighth production block. Of these children, three were in the consistent-print condition and two were in the no-print condition. In addition, data for two children were excluded from the analyses because their irregular word reading performance was 2 standard deviations below the sample mean. As shown in Table 1, children across conditions did not differ on reading and age (ps > .64) or on parent education and family income (ps > .71). Importantly, children across studies and conditions did not differ on these variables (ps > .18, median p = .78) with one exception, namely that families in Study 1 reported, on average, a higher income (p > .001) than those in Study 2.

Receptive vocabulary As shown in the top panel of Fig. 2, bilingual children in the two print conditions (i.e., consistent print and inconsistent print) had less difficulty in repeating novel labels during training as compared with children in the no-print condition. In fact, children’s performance in both print conditions was at ceiling. Given this, no statistical analyses were performed on these data. The top panel of Fig. 2 also shows that posttest recognition performance was near or at ceiling, indicating that all children had learned an association between the label and its referent. The finding that children in the no-print condition were also able to recognize more labels on posttest than they were able to repeat accurately during training raises the possibility that exposure along with opportunities to repeat labels (albeit not necessarily accurately) was sufficient for these children to make some gains in receptive vocabulary. The tests of expressive vocabulary conducted next provide information on the accuracy of the phonological representations encoded during training. Expressive vocabulary The bottom panel of Fig. 2 illustrates children’s accuracy in producing spoken labels during the learning phase. An ANCOVA with regular word reading and home language as covariates revealed a significant main effect of condition, F(2, 57) = 16.65, MSE = 10.15, p < .001, g2 = .37. The print facilitation contrast was significant, indicating that children in the two print conditions produced more labels accurately during learning than children in the no-print condition, contrast estimate = 1.56, CI = [0.37, 2.15], p < .001, Cohen’s d = 1.69. Moreover, the print consistency contrast was also significant, contrast estimate = 0.68, CI = [–0.01, –1.36], p = .05, Cohen’s d = –0.49. Counter to predictions, however, the incidental presence of inconsistent print facilitated production more than the presence of consistent print. In addition, there was a Cycle  Condition interaction, F(16, 456) = 5.01, MSE = 0.89, p = .001, g2 = .15. Follow-up polynomial contrasts were examined for each condition separately (Bonferroni-corrected

257

ax = 6

K. Jubenville et al. / Journal of Experimental Child Psychology 126 (2014) 245–263

ax = 6

Posest recognion

Posest recall

Fig. 2. Mean numbers of nonwords as a function of learning condition in Study 2. Top panel: Nonwords repeated during training and recognized at posttest. Bottom panel: Nonwords produced during training and recalled at posttest. Error bars represent standard errors.

p < .02). As in Study 1, there were significant linear trends in all conditions (ps < .001), but the significant quadratic trends were limited to the two print conditions (ps < .001). These results show that children were gradually learning in all conditions and that for the two print conditions the sharpest increases were in the initial trials and the smallest increases were in the final trials. No other contrast was significant. Fig. 2 also presents posttest recall performance results. The analysis of variance (ANOVA) yielded a main effect of condition, F(2, 59) = 11.16, MSE = 2.64, p < .001, g2 = .27. The print facilitation contrast was significant, showing that children in the two print conditions recalled more spoken labels than children in the no-print condition, contrast estimate = 1.77, CI = [0.91, 2.64], p < .001, Cohen’s d = 1.23. Here also, the effect of print consistency was reversed; it was the inconsistent print that yielded stronger recall as opposed to consistent print, contrast estimate = –1.20, CI = [–0.17, –2.23], p = .02, Cohen’s d = –0.64. An interpretation of this unexpected finding is found later in the General Discussion.

258

K. Jubenville et al. / Journal of Experimental Child Psychology 126 (2014) 245–263

Spelling A MANOVA for posttest spelling of the first and second syllables yielded a main effect of condition, Wilks’ k(4, 116) = .375, p < .001. As in Study 1, children in the two print conditions had no difficulty in spelling the first syllable (Ms = 5.8 and 5.7, SDs = 0.6 and 0.6, for the consistent-print and inconsistentprint conditions, respectively). As expected, the difference appeared on the second syllable; children in the consistent-print condition spelled more items accurately (M = 5.4, SD = 0.9) than children in the inconsistent-print condition (M = 3.5, SD = 1.2), contrast estimate = 1.85, CI = [1.17, 2.53], p = .001, Cohen’s d = 1.79. Children in the no-print condition spelled the majority of nonwords as they were written in the consistent-print condition (Ms = 3.9 and 4.3, SDs = 1.6 and 1.0, for the first and second syllables, respectively). We now turn to the issue of whether phonology could help to construct accurate orthographic representations for bilinguals by examining second-syllable errors made by bilingual children in the inconsistent-print condition. Fully 93% of second-syllable errors were on the silent letter, with the majority (55%) of these consisting of substituting the silent consonant with other consonants or a silent e and the remaining consisting of omissions. As in Study 1, these spelling errors highlight the interplay between phonology and orthography. Supplemental analyses The issue of accurate phonological representations was analyzed further by examining the stability of children’s production errors during learning as well as on subsequent posttests. These error analyses replicated the findings of Study 1. Of the errors made when producing a target item on the last three production cycles during learning, 26% were stable (i.e., identical) erroneous representations in the no-print condition versus 7% in the two print conditions combined. On posttest, the 17 bilingual children with stable erroneous representations in the no-print condition tended to recall (72%) and spell (81%) these erroneous representations. Only six children across the two print conditions had erroneous representations, and they tended to recall (83%) but not spell (33%) these erroneous representations. The greater number of children with stable erroneous phonological representations in the no-print condition provides further evidence that orthography can help to clarify and maintain phonological representations in memory. Next, we assessed the relation between labels produced during training and at posttest by calculating the probability of recalling a label as a function of the number of times it was produced at training. This item analysis revealed a gradual steplike increase in recall probability that was similar across conditions. For words produced from one to eight times during training, the recall probabilities were .18, .46, .51, .52, .58, .75, .96, and .91, respectively. In contrast, labels that were not produced during training had only a .08 probability of being recalled. This analysis suggests that opportunities to recall novel labels are linked to gains in expressive vocabulary. Children in this study spoke a variety of languages; therefore, it was necessary to verify that this did not affect the findings. To do so, the performance of French–English bilinguals (n = 40) was contrasted with that of the remaining bilingual and multilingual children (n = 22). Importantly for the current study, children’s mean performances across conditions for these two groups showed identical patterns for receptive and expressive vocabulary as well as spelling. General discussion In the current research, we examined the interaction between phonological and orthographic information in an oral word-learning paradigm. In two between-participant studies, we tested the theoretical claim that orthography, even when presented incidentally, facilitates the acquisition of accurate and precise phonological representations. With language and reading experience, phonological and orthographic representations become bonded in the mental lexicon and, therefore, increase the quality of word form representations as well as the availability of their contents (Ehri, 1978; Perfetti & Hart, 2002; Share, 1999). This theoretical prediction was supported in two studies; both monolingual and bilingual children who were exposed to print incidentally produced more novel spoken labels during training and delayed recall than children who never saw the words’ spelling. A second goal

K. Jubenville et al. / Journal of Experimental Child Psychology 126 (2014) 245–263

259

of the research was to assess whether the consistency of print would moderate this facilitative effect. We did find moderating effects; however, they differed across studies. In Study 1 it was the monolingual children in the consistent-print condition who produced the most novel labels, whereas in Study 2 it was the bilingual children in the inconsistent-print condition who produced and recalled the most labels. The facilitative effect of print and the moderating effect of print consistency on oral vocabulary acquisition are discussed below. Orthographic facilitation The incidental presence of print during learning increased children’s acquisition of expressive vocabulary as compared with children who learned without print, replicating past results (Lucas & Norbury, 2013; Mengoni et al., 2013; Ricketts et al., 2009; Rosenthal & Ehri, 2008). The facilitative effect of print was strong for both monolingual children in Study 1 (Cohen’s d = 1.55) and bilingual children in Study 2 (Cohen’s d = 1.69). These findings support the claim that exposure to print can help to establish accurate phonological representations. Importantly, we did not find evidence that the print facilitation effect would be attenuated due to bilingualism. That is, the documented retrieval weaknesses of bilinguals (Bialystok et al., 2008) did not hinder performance in the current studies. In practice, this implies that showing literate children the spelling for a word they are learning orally might help them to acquire it more quickly even if no attention is drawn to it. Additional support for the claim that the incidental presence of print facilitates the acquisition of accurate phonological representations was obtained by examining children’s production errors during learning. In both studies, the majority of children in the no-print condition established at least one stable erroneous pronunciation of the novel words (i.e., 63% and 74% of children in Studies 1 and 2, respectively). Once these erroneous pronunciations were established, children tended to maintain them at recall or produced another erroneous pronunciation. Note, however, that the presence of print did not guarantee accuracy given that a small number of children (4 and 7 in Studies 1 and 2, respectively) also established erroneous representations that were maintained at recall. In contrast to previous research, children in the no-print condition did not reach ceiling on repetition trials during learning (Ricketts et al., 2009). In fact, they repeated on average two words less than children in the two print conditions (top panel in Figs. 1 and 2). Consequently, it might be argued that the facilitative effect of print on expressive vocabulary learning was due to children in the no-print condition having received less exposure to the targets. The learning procedure, however, circumvented this possibility because children in the no-print condition heard each word twice as frequently during training as compared with children in the two print conditions. A second argument might be that children in the no-print condition were not learning to match the novel words to their pictured referents. This was clearly not the case; children in the no-print condition performed at ceiling on the receptive vocabulary posttest. The discrepancy, in the no-print condition, between the repetition performance at training and the receptive vocabulary on posttest suggests that accurate and precise phonological representations might not be necessary for receptive vocabulary. Presumably, receptive vocabulary requires a sufficient, but not necessarily complete, phonological overlap between the label heard and the representation stored in memory. This interesting possibility could be examined in future studies that include a greater number of stimuli to avoid ceiling effects on posttest performance. Comparing children’s repetition and production performance highlights the differences between learning receptive and expressive vocabulary. Our findings are consistent with those of Sénéchal (1997), who found that exposure to the to-be-learned items was sufficient for the acquisition of receptive vocabulary but that producing the words during learning facilitated expressive vocabulary acquisition. Although Sénéchal’s study was conducted with preschoolers, the current findings suggest that the differences in learning persist in elementary school. Orthographic consistency In addition to demonstrating that print can facilitate oral word learning, it is important to understand what factors might influence the effectiveness of orthography as an ‘‘anchor’’ for pronunciations

260

K. Jubenville et al. / Journal of Experimental Child Psychology 126 (2014) 245–263

in memory. Because the locus of this effect is believed to lie in an individual’s ability to exploit phoneme–grapheme correspondences, the lack of congruency between phonological and orthographic forms in opaque languages should moderate the extent to which spelling can act as a memory aid for pronunciations. Indeed, the obtained results showed that the predictability of mappings from speech to print affected children’s learning and posttest recall. The effect, however, was more complex than anticipated. In Study 1, children who were exposed to consistent print incidentally produced more novel words during training as compared with those who saw inconsistent print (Cohen’s d = 0.98). This finding replicates the results of Ricketts and colleagues (2009), who found consistency effects during training. However, we did not find a consistency effect on delayed recall; consequently, our results circumscribe those of Ricketts and colleagues by showing that a consistency effect might be short-lived—at least for monolingual children. In contrast to Study 1, the performance of bilingual children in Study 2 showed the reverse pattern—an advantage for children who learned words with inconsistent spellings rather than the predicted advantage for children exposed to consistent spellings (Cohen’s d = 0.68). There are at least two complementary explanations for this effect. One possible explanation is that children paid greater attention to the inconsistent words. Recall that the source of the inconsistency was a silent consonant at the end of nonwords and that this type of inconsistency is seldom, if ever, found in other alphabetic languages. Previous findings on automatic activation of orthographic knowledge in bilinguals (Jared & Kroll, 2001) suggests that, on hearing the label, bilingual children might have paid greater attention to the print in an effort to resolve the mismatch between the French orthography and the orthography of the other languages spoken. Some support for this possibility comes from children’s spelling errors; whereas monolingual children most frequently omitted the silent consonant, bilingual children most frequently substituted the silent consonant with another letter. Future research could exploit the use of eye-tracking software to allow for a comparison of the degree to which children in the two print conditions attend to the print. Another possible explanation for the inconsistent-print advantage found in Study 2 concerns children’s cognitive control and metalinguistic skills. There is some evidence that bilingual children have superior cognitive control and metalinguistic skills than monolingual children (e.g., Bialystok, 2008; Laurent & Martinot, 2010). These superior skills in bilingual children might result in their encoding the orthography of inconsistent labels with more precision (i.e., encoding the final silent consonant). Some support for this interpretation is found in children’s spelling performance; bilingual children in the inconsistent-print condition spelled 58% of the syllables with silent consonants accurately, as compared with 43% for monolingual children. This plausible explanation, however, is speculative. Future research could include measures of executive function and metalinguistic skills to examine this possibility further. A minority (17%) of children in Study 2 was multilingual, and although their pattern of performance across conditions was similar to that of the French–English bilinguals, it will be important to replicate the obtained inconsistency effect with a more constrained sample. For instance, future research could limit the sample to children speaking the same two languages. In addition, children’s level of bilingualism could be assessed directly with measures of vocabulary as well as phonological and orthographic processing in order to understand this intriguing effect. In the current research, the use of nonwords ensured greater control because the stimuli across conditions were matched on the pronunciation of the final syllable and on the number of phonological and orthographic neighbors for that final syllable. Moreover, it ensured that all stimuli were unknown and that they were equated on number of phonemes, letters, and syllables. Importantly, we replicated with nonwords the print facilitation effect found by Rosenthal and Ehri (2008), who used real rare words, as well as Lucas and Norbury (2013), who used science words. This being said, it will be important to examine the consistency effect with real words. In a similar vein, we measured children’s receptive knowledge of the new labels with a recognition task and their expressive knowledge with a recall task at posttest. These tasks, constructed like existing tests of receptive and expressive vocabulary, tapped children’s associative knowledge between a referent and its label. Future research could also examine whether the print consistency effect is present in a learning paradigm that includes definitions for words, as was done in Rosenthal and Ehri’s (2008) research on print facilitation.

K. Jubenville et al. / Journal of Experimental Child Psychology 126 (2014) 245–263

261

Given the interest in orthographic processing in the current research, children were matched on word reading across conditions, and these triplets were assigned randomly to the conditions. Future research might also include additional control variables such as vocabulary knowledge, phonological processing, and nonverbal ability. At the same time, it is noteworthy that we replicated and extended the findings of Ricketts and colleagues (2009), who controlled for vocabulary and nonverbal ability. It is also of interest to highlight the findings of Lucas and Norbury (2013), showing that even though the typically developing children in their study had a greater vocabulary size than children with autism, both groups learned equally well when print was present. In the current research, the control stimuli consisted of a picture with no additional cue. Hence, one could question whether the print facilitation that we obtained was due to orthography per se as opposed to any additional cue. Three studies, however, clearly showed that the print facilitation effect is due to the presence of orthography and not an additional cue (Ehri & Wilce, 1979; Hu, 2008; Mengoni et al., 2013). Implicit learning of orthography The current research extended to French the previous findings showing that literate children acquired orthographic representations implicitly from incidental exposure to print (Ricketts et al., 2009; Rosenthal & Ehri, 2008, 2011). Children exposed incidentally to nonwords spelled, on average, more nonwords accurately than children who never saw print. This difference was not due to children in the no-print condition demonstrating spelling difficulty but rather was due to their frequent use of legal French homophonic spelling alternatives (e.g., bonkli instead of boncli) or to spelling their erroneous phonological representation of the novel words (e.g., noudri instead of loudri). This implies that incidental exposure to orthography clarified word-specific phonology that supports the claim that orthography had become amalgamated with phonology in memory, resulting in more precise lexical knowledge. The current research also extended to French the previous findings of phoneme-to-grapheme consistency influencing spelling accuracy in English (Ricketts et al., 2009). Children in the consistent-print condition in Study 1 (Cohen’s d = 2.67) and Study 2 (Cohen’s d = 1.79) demonstrated more accurate memories for orthographic forms than children in the inconsistent-print condition, indicating that while children were implicitly acquiring orthography, the resulting representation was more precise when spelling was more predictable. Consistent with past findings, more than 90% of spelling errors in the inconsistent-print condition occurred on the silent consonant endings (Sénéchal et al., 2006). Such findings are consistent with the notion that orthographic representations are initially constructed as a ‘‘frame’’ based on the clearly specified (i.e., consistent) graphemes, whereas the inconsistent portions of words are not represented or are maintained by underspecified ‘‘spacers’’ (McKague, Davis, Pratt, & Johnston, 2008). It equally reflects challenges faced by spellers of opaque languages, with French simply being the example used in the current studies. In light of the finding that orthography can be acquired implicitly, and evidence of a frame being formed in memory even for inconsistent words, it follows that displaying a word’s spelling while it is being taught orally may provide additional benefits to users of opaque orthographies. In sum, the current research extended previous findings of orthographic facilitation from English to another opaque language—French. It also made an attempt to establish parameters around this effect through a controlled investigation of phoneme-to-grapheme consistency. Finally, it extended previous research by examining performance in both monolingual and bilingual children. Overall, results support the claim that expressive vocabulary acquisition is influenced not only by ear but also by eye. Acknowledgments This research was funded by the Social Sciences and Humanities Research Council of Canada. We sincerely thank the children and their parents for their willingness to participate; the teachers and school staff for facilitating data collection; and Amy Winchester, Yannick Bicamumpaka, and Maxime Gingras for their invaluable assistance with this project.

262

K. Jubenville et al. / Journal of Experimental Child Psychology 126 (2014) 245–263

Appendix Orthography, pronunciation, and numbers of orthographic and phonological neighbors for nonwords as a function of phoneme-to-grapheme consistency.

Consistent

Inconsistent

Orthography

ON

Pronunciation

PN

Orthography

ON

Pronunciation

PN

boncli poncra moubla loudri reuglo teuvro

0 3 2 1 0 0

/bO~kli/ /pO~ka/ /mubla/ /ludi/ /øglo/ /tøvo/

0 1 2 0 0 0

boclid pocrat mublap ludrit reglot tevrop

0 3 1 1 0 0

/bokli/ /poka/ /mybla/ /lydi/ /Eglo/ /tEvo/

0 0 0 0 0 0

Note. ON, orthographic neighbors; PN, phonological neighbors.

References Aud, S., Hussar, W., Kena, G., Bianco, K., Frohlich, L., Kemp, J., et al (2011). The condition of education 2011 (NCES 2011-033, National Center for Education Statistics). Washington, DC: Government Printing Office. Baayen, H. R., Piepenbrock, R., & Gulikers, L. (1995). The CELEX lexical database, Release 2 [CD-ROM]. Philadelphia: University of Pennsylvania, Center for Lexical Information. Bialystok, E. (2008). Consequences of bilingualism for cognitive development. In J. R. Kroll & A. de Groot (Eds.), Handbook of bilingualism: Psycholinguistic approaches (pp. 417–432). Oxford, UK: Oxford University Press. Bialystok, E., Craik, F., & Luk, G. (2008). Cognitive control and lexical access in younger and older bilinguals. Journal of Experimental Psychology: Learning, Memory, and Cognition, 34, 859–873. Content, A., Mousty, P., & Radeau, M. (1990). BRULEX: Une base de données lexicales informatisée pour le Français écrit et parlé. L’Année Psychologique, 90, 551–566. http://dx.doi.org/10.3406/psy.1990.29428. Deacon, S. H., Wade-Woolley, L., & Kirby, J. R. (2009). Flexibility in young second-language learners: Examining the language specificity of orthographic processing. Journal of Research in Reading, 32, 215–229. Ehri, L. C. (1978). Beginning reading from a psycholinguistic perspective: Amalgamation of word identities. In F. B. Murray (Ed.), The recognition of words (pp. 1–33, IRA Series on the Development of the Reading Process, No. 3). Newark, DE: International Reading Association. Ehri, L. C., & Wilce, L. (1979). The mnemonic value of orthography among beginning readers. Journal of Educational Psychology, 71, 26–40. Haahr, M. (1998). Random Generator [online]. http://www.random.org. Hu, C.-F. (2008). Use orthography in L2 auditory word learning: Who benefits? Reading and Writing, 21, 823–841. Jared, D., Cormier, P., Levy, B. A., & Wade-Woolley, L. (2011). Early predictors of biliteracy development in children in French immersion: A 4-year longitudinal study. Journal of Educational Psychology, 103, 119–139. Jared, D., Cormier, P., Levy, B. A., & Wade-Woolley, L. (2013). Discrimination of English and French orthographic patterns by biliterate children. Journal of Experimental Child Psychology, 114, 469–488. Jared, D., & Kroll, J. F. (2001). Do bilinguals activate phonological representations in one or both of their languages when naming words? Journal of Memory and Language, 44, 2–31. Katz, L., & Frost, R. (1992). The reading process is different for different orthographies: The orthographic depth hypothesis. In R. Frost & L. Katz (Eds.), Advances in psychology. Orthography, phonology, morphology, and meaning (Vol. 94, pp. 67–84). Amsterdam: Elsevier North-Holland. Laurent, A., & Martinot, C. (2010). Bilingualism and phonological awareness: The case of bilingual (French–Occitan) children. Reading and Writing, 23, 435–452. Lucas, R., & Norbury, C. F. (2013). Orthography facilitates vocabulary learning for children with autism spectrum disorders (ASD). Quarterly Journal of Experimental Psychology. http://dx.doi.org/10.1080/17470218.2013.859714. Advance online publication. McKague, M., Davis, C., Pratt, C., & Johnston, M. B. (2008). The role of feedback from phonology to orthography in orthographic learning: An extension of item-based accounts. Journal of Research in Reading, 31, 55–76. Mengoni, S. E., Nash, H., & Hulme, C. (2013). The benefit of orthographic support for oral vocabulary learning in children with Down syndrome. Journal of Child Language, 40, 221–243. Mousty, P., Leybaert, J., Alegria, J., Content, A., & Morais, J. (1994). BELEC: Batterie d’Évaluation du Langage Écrit et de ses Troubles. In J. Grégoire & B. Piérart (Eds.), Evaluer les troubles de la lecture: Les nouveaux modèles théoriques et leurs implications diagnostiques (pp. 127–145). Bruxelles, Belgium: De Boeck. Peereman, R., & Content, A. (1999). LEXOP: A lexical database providing orthography–phonology statistics for French monosyllabic words. Behavioral Research Methods, Instruments, & Computers, 31, 376–379. Peereman, R., Sprenger-Charolles, L., & Messaoud-Galusi, S. (2013). The contribution of morphology to the consistency of spelling to sound relations: A quantitative analysis based on French elementary school readers. L’Année Psychologique [Topics in Cognitive Psychology], 113, 3–33. http://dx.doi.org/10.4074/S0003503313001012.

K. Jubenville et al. / Journal of Experimental Child Psychology 126 (2014) 245–263

263

Perfetti, C. A., & Hart, L. (2002). The lexical quality hypothesis. In L. Verhoeven, C. Elbro, & P. Reitsma (Eds.). Precursors of functional literacy (Vol. 11, pp. 189–213). Amsterdam: John Benjamins. Perry, C., Ziegler, J., & Zorzi, M. (2014). When silent letters say more than a thousand words: An implementation and evaluation of CDP++ in French. Journal of Memory and Language, 72, 98–115. Ricketts, J., Bishop, D. V. M., & Nation, K. (2009). Orthographic facilitation in oral vocabulary acquisition. Quarterly Journal of Experimental Psychology, 62, 1948–1966. Rosenthal, J., & Ehri, L. C. (2008). The mnemonic value of orthography for vocabulary learning. Journal of Educational Psychology, 100, 175–191. Rosenthal, J., & Ehri, L. C. (2011). Pronouncing new words aloud during the silent reading of text enhances fifth graders’ memory for vocabulary words and their spellings. Reading and Writing, 24, 921–950. Roux, S., & Bonin, P. (2013). ‘‘With a little help from my friends’’: Orthographic influences in spoken word recognition. L’année Psychologique [Topics in Cognitive Psychology], 113, 35–48. http://dx.doi.org/10.4074/S0003503313001024. Sénéchal, M. (1997). The differential effect of storybook reading on preschoolers’ acquisition of expressive and receptive vocabulary. Journal of Child Language, 24, 123–138. Sénéchal, M. (2006). Testing the home literacy model: Parent involvement in kindergarten is differentially related to Grade 4 reading comprehension, fluency, spelling, and reading for pleasure. Journal for the Scientific Study of Reading, 10, 59–87. Sénéchal, M., Basque, M. T., & Leclaire, T. (2006). Morphological knowledge as revealed in children’s spelling accuracy and reports of spelling strategies. Journal of Experimental Child Psychology, 95, 231–254. Share, D. L. (1999). Phonological recoding and orthographic learning: A direct test of self-teaching hypothesis. Journal of Experimental Child Psychology, 29, 294–305. Statistics Canada. (2007). 2006 Community profiles, 2006 census (Catalogue No. 92-591-XWE). Ottawa, Ontario: Author. Available at: . Statistics Canada. (2012). Census profile, 2012 census (Statistics Canada Catalogue No. 98-316-XWE). Ottawa, Ontario: Author. Available at: . United Nations Educational, Scientific, and Cultural Organisation (2003). Education in a multilingual world (ED-2003/WS/2). Paris: UNESCO. Zarei, A. A. (2009). The effect of bimodal, standard, and reversed subtitling on L2 vocabulary recognition and recall. Journal of the Society for Information Display, 49, 65–85.

The moderating effect of orthographic consistency on oral vocabulary learning in monolingual and bilingual children.

Two studies were conducted to assess whether (a) the incidental presence of print facilitates the acquisition of oral vocabulary, (b) the facilitative...
830KB Sizes 3 Downloads 2 Views