Do eye movements reveal differences between monolingual and bilingual children’s first-language and second-language reading? A focus on word frequency effects

Journal of Experimental Child Psychology 173 (2018) 318–337

Contents lists available at ScienceDirect

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

Do eye movements reveal differences between monolingual and bilingual children’s first-language and second-language reading? A focus on word frequency effects Veronica Whitford a,⇑, Marc F. Joanisse b a b

Department of Psychology, University of Texas at El Paso, El Paso, Texas 79902, USA Department of Psychology, Brain and Mind Institute, University of Western Ontario, London, Ontario N6A 3K7, Canada

a r t i c l e

i n f o

Article history: Received 18 August 2017 Revised 26 November 2017

Keywords: Reading Eye movements Word frequency Children Bilingualism Monolingualism

a b s t r a c t An extensive body of research has examined reading acquisition and performance in monolingual children. Surprisingly, however, much less is known about reading in bilingual children, who outnumber monolingual children globally. Here, we address this important imbalance in the literature by employing eye movement recordings to examine both global (i.e., text-level) and local (i.e., word-level) aspects of monolingual and bilingual children’s reading performance across their first-language (L1) and secondlanguage (L2). We also had a specific focus on lexical accessibility, indexed by word frequency effects. We had three main findings. First, bilingual children displayed reduced global and local L1 reading performance relative to monolingual children, including larger L1 word frequency effects. Second, bilingual children displayed reduced global and local L2 versus L1 reading performance, including larger L2 word frequency effects. Third, both groups of children displayed reduced global and local reading performance relative to adult comparison groups (across their known languages), including larger word frequency effects. Notably, our first finding was not captured by traditional offline measures of reading, such as standardized tests, suggesting that these measures may lack the sensitivity to detect such nuanced between-group differences in reading performance. Taken together, our findings demonstrate that bilingual children’s simultaneous exposure to two reading systems

⇑ Corresponding author at: Department of Psychology, University of Texas at El Paso, 500 West University Avenue, El Paso, Texas 79902, USA. Tel.: +1 915 747 7514; Fax: +1 915 747 6553. E-mail address: [email protected] (V. Whitford). https://doi.org/10.1016/j.jecp.2018.03.014 0022-0965/Ó 2018 Elsevier Inc. All rights reserved.

V. Whitford, M.F. Joanisse / Journal of Experimental Child Psychology 173 (2018) 318–337

319

leads to eye movement reading behavior that differs from that of monolingual children and has important consequences for how lexical information is accessed and integrated in both languages. Ó 2018 Elsevier Inc. All rights reserved.

Introduction Reading is arguably one of the most important neurocognitive skills that children learn. Indeed, it is strongly linked to their academic success (e.g., La Paro & Pianta, 2000) and, ultimately, to their economic, occupational, and social success in later years (e.g., Green & Riddell, 2007; Kirsch, Jungeblat, Jenkins, & Kolstad, 2002). Given its centrality to nearly all domains of modern life, it is no surprise that a rich body of literature has investigated the perceptual, oculomotor, cognitive, and linguistic processes implicated in reading through the use of online reading measures, most notably eyetracking, which offers a direct, naturalistic, and temporally precise measure of these processes (reviewed in Rayner, 1998, 2009; Rayner, Pollatsek, Ashby, & Clifton, 2012; Whitford, Pivneva, & Titone, 2016). In turn, this literature has given rise to several well-formulated computational models that can account for these key reading processes, such as EZ Reader (e.g., Pollatsek, Reichle, & Rayner, 2006; Reichle, Pollatsek, Fisher, & Rayner, 1998) and SWIFT (e.g., Engbert, Nuthmann, Richter, & Kliegl, 2005). Although this work has proved crucial in advancing our knowledge and understanding of reading behavior, its primary focus has been on skilled reading in university-aged young adults. Thus, less is known about reading behavior in children for whom many of these key reading processes are still developing. The relatively small but growing developmental eye movement reading literature has reported largely quantitative differences in the eye movement record of typically developing children versus young adults (reviewed in Blythe & Joseph, 2011; Frey, 2016; Rayner, 1998, 2009; Rayner et al., 2012; Reichle et al., 2013). These differences include more fixations, longer fixation durations, less skipping, more saccades (both progressive and regressive), and shorter saccade amplitudes in children that, collectively, culminate in slower overall reading rates. Children also show reduced parafoveal processing, including a smaller attentional or perceptual span, which reflects the amount of useful visual information obtained during fixation (e.g., Häikiö, Bertram, & Hyönä, 2010; Häikiö, Bertram, Hyönä, & Niemi, 2009; Rayner, 1986; Sperlich, Meixner, & Laubrock, 2016; Sperlich, Schad, & Laubrock, 2015; Tiffin-Richards & Schroeder, 2015). These differences, however, decrease as children’s reading skills improve with age (e.g., Ducrot, Pynte, Ghio, & Lété, 2013; Huestegge, Radach, Corbic, & Huestegge, 2009; Leeuw, Segers, & Verhoeven, 2015; McConkie et al., 1991; Vorstius, Radach, & Lonigan, 2014), with some reports suggesting that their eye movements and perceptual span pattern with those of young adults at approximately 11 or 12 years old (e.g., Blythe & Joseph, 2011; Rayner, 1986; Reichle et al., 2013). Age differences in the eye movement record likely reflect some combination of children’s developing linguistic knowledge and more peripheral visual and oculomotor processes (e.g., Liang, Wang, Yang, & Bai, 2017; Luke, Henderson, & Ferreira, 2015; Mancheva et al., 2015; Reichle et al., 2013). As children’s language proficiency (across sublexical, lexical, syntactic, and discourse levels) increases through continued exposure and/or as their oculomotor system further develops, their eye movements likely become more fine-tuned, resulting in adult-like reading behavior. Consistent with this conjecture, recent eye movement reading research has found that children differ from young adults in terms of pragmatic (Joseph et al., 2008) and syntactic (Joseph & Liversedge, 2013) processing, as well as in their sensitivity to key linguistic variables known to affect lexical processing, such as word length and word frequency (reviewed in Blythe & Joseph, 2011; Frey, 2016; Rayner, 1998, 2009; Rayner et al., 2012; Reichle et al., 2013). With respect to word length, studies have found that both children and young adults fixate longer words more often and for more time than shorter words (e.g., refrigerator vs. stove). However, these

320

V. Whitford, M.F. Joanisse / Journal of Experimental Child Psychology 173 (2018) 318–337

effects are reliably larger in children and are driven by their differentially slower processing of longer words (Aghababian & Nazir, 2000; Blythe, Häikiö, Bertam, Liversedge, & Hyönä, 2011; Huestegge et al., 2009; Hyönä & Olson, 1995; Joseph, Liversedge, Blythe, White, & Rayner, 2009; Luke et al., 2015; Tiffin-Richards & Schroeder, 2015; Vitu, McConkie, Kerr, & O’Regan, 2001). Similarly, both children and young adults exhibit word frequency effects, whereby they fixate lower-frequency words more often and for more time than higher-frequency words (e.g., hutch vs. table). However, there is some evidence that these effects are larger in children and are driven by their differentially slower processing of lower-frequency words (Blythe, Liversedge, Joseph, White, & Rayner, 2009; Huestegge et al., 2009; Hyönä & Olson, 1995; Joseph, Nation, & Liversedge, 2013; Luke et al., 2015; Tiffin-Richards & Schroeder, 2015; Vorstius et al., 2014; cf. Blythe et al., 2006). Given that word frequency effects provide a useful and well-studied proxy of how individuals mentally represent and access words during recognition, this latter finding suggests that children’s inherently reduced experience with print (and language more generally) may lead to reduced lexical accessibility and, consequently, more effortful reading. Indeed, recent modeling work has demonstrated that children’s reduced rates of lexical processing are a key determinant of their eye movement behavior during reading (see Mancheva et al., 2015; Reichle et al., 2013). To summarize so far, eye movement studies have reported reduced reading performance in typically developing children versus young adults, which likely reflects their reduced ease of word processing. Although this work has provided important insights into the developmental trajectory of eye movement reading behavior, it has almost exclusively focused on reading in monolingual children. To date, only one study has used eye movements to examine reading performance in bilingual children and found that second-language (L2) speakers of Spanish were less successful at using cues to process grammatical gender than their first-language (L1) speaking peers (Lew-Williams, 2017). We note, however, that numerous studies have examined bilingual children’s reading and readingrelated skills through a variety of other tasks, although often offline in nature. In general, this work has found that word-level reading skills are comparable in monolingual and bilingual children, especially when bilingual children acquire their L2 early in life and/or as they age. In contrast, text-level reading skills are often variable between these two groups (reviewed in August & Shanahan, 2006; Jared, 2015). This disparity in the developmental eye movement reading literature is surprising, given that childhood bilingualism is an ever-increasing global phenomenon. Moreover, in light of recent work demonstrating that knowledge and use of more than one language affect eye movement measures of L2 and, more interestingly, L1 reading performance in bilingual younger (Cop, Drieghe, & Duyck, 2015; Cop, Keuleers, Drieghe, & Duyck, 2015; Whitford & Titone, 2012, 2015, 2016, 2017a,b) and older (Titone, Whitford, Lijewska, & Itzhak, 2016; Whitford & Titone, 2016, 2017a,b) adults, bilingual children’s simultaneous exposure to two reading systems may lead to differences in the eye movement record. Thus, the findings from eye movement studies involving monolingual children might not generalize to bilingual children. The current study addressed this important imbalance in the literature by examining both global (i.e., text-level) and local (i.e., word-level) aspects of monolingual and bilingual children’s L1 and L2 reading behavior (indexed by fixation and saccadic measures), with a specific focus on lexical accessibility (indexed by word frequency effects). To ensure ecological relevance to reading in classroom environments, naturalistic texts were used while children’s eye movements were monitored. Groups of monolingual and bilingual young adults were also included to examine developmental differences in the eye movement record. Theoretical motivation for the current study comes from a bilingual adaptation of Perfetti’s lexical quality hypothesis (e.g., Perfetti, 2007; Perfetti & Hart, 2002) known as the weaker links hypothesis (Gollan, Montoya, Cera, & Sandoval, 2008; Gollan et al., 2011). As a lexical entrenchment account of word processing, the weaker links hypothesis proposes that continued exposure to words strengthens their orthographic, phonological, and semantic links, leading to greater lexical quality and accessibility (i.e., more entrenched lexical representations) and, ultimately, smaller word frequency effects (for similar accounts, see also Andrews & Hersch, 2010; Dijkstra & Van Heuven, 2002; Joanisse & McClelland, 2015; Kuperman & Van Dyke, 2013; Monaghan, Chang, Welbourne, & Brysbaert, 2017; Seidenberg & McClelland, 1990). The effect of continued exposure on ease of

V. Whitford, M.F. Joanisse / Journal of Experimental Child Psychology 173 (2018) 318–337

321

word processing may, however, reach a functional ceiling over time. As such, lower-frequency words, whose lexical representations are far from ceiling, may benefit more from continued exposure than higher-frequency words, whose lexical representations are near or at ceiling. Moreover, continued (or greater) exposure to all words may reduce the magnitude of word frequency effects, as differences in lexical quality and accessibility between lower-frequency and higher-frequency words would be reduced. Accordingly, the weaker links hypothesis makes the following three key predictions. First, weaker links in bilinguals versus monolinguals; bilinguals have less absolute exposure to words than monolinguals (given their divided language knowledge and use) and, thus, should exhibit larger word frequency effects. Second, weaker L2 versus L1 links among bilinguals; bilinguals generally have less absolute exposure to L2 words than to L1 words and, thus, should exhibit larger L2 word frequency effects. Third, weaker links in children versus adults; children have less absolute exposure to words than adults (because their language and literacy skills are still developing) and, thus, should exhibit larger word frequency effects.1 Across all three predictions, larger word frequency effects should be driven by differentially slower processing of lower-frequency words, as their lexical representations are further from a functional ceiling than those of higher-frequency words. Recent eye movement work examining reading in bilingual adults has provided some empirical support for the weaker links hypothesis. For instance, Whitford and Titone (2012, 2017a,b) found that both English–French bilingual younger and older adults experience reduced L2 versus L1 lexical quality and accessibility during extended paragraph reading; this finding was indexed by larger L2 versus L1 word frequency effects and driven by slower processing of lower-frequency L2 words (see also Cop, Keuleers et al., 2015, for similar effects during novel reading in young adults). Whitford and Titone (2015, 2016) also found that such reductions in lexical entrenchment scale up to affect sentence-level measures of L2 versus L1 reading performance in different samples of English–French bilingual younger and older adults; this finding was indexed by slower reading rates, longer fixation durations, shorter progressive saccades, and more regressive saccades during L2 reading (see also Cop, Drieghe et al., 2015, for similar effects during novel reading in young adults). Interestingly, however, eye movement studies comparing reading in monolingual and bilingual young adults have found no compelling evidence of weaker links. For instance, Cop, Keuleers et al. (2015) found comparable word frequency effects during L1 novel reading in English monolinguals and Dutch–English bilinguals (see also Gollan et al., 2011, for comparable word frequency effects in English monolinguals, Dutch–English bilinguals, and Spanish–English bilinguals, albeit during L2 sentence reading). In a reanalysis of their data, Cop, Drieghe et al. (2015) also found comparable L1 sentence-level reading behavior in these two groups. Because their monolingual and bilingual participants had comparable levels of L1 proficiency (and, by proxy, comparable levels of L1 exposure), their findings suggest that similar amounts of L1 experience can offset weaker links. As mentioned earlier, only one eye movement study has examined reading performance in bilingual children; it investigated how two groups of bilingual children processed grammatical gender in their L1 and L2, respectively (Lew-Williams, 2017). As such, an open question is whether weaker links also extend to reading in bilingual children (across their known languages). In accordance with the weaker links hypothesis, the following three predictions are made for the current study. First, bilingual children will exhibit reduced global (i.e., text-level) and local (i.e., world-level) L1 reading performance compared with monolingual children; these reductions will also include larger L1 word frequency effects. Second, bilingual children will exhibit reduced global and local L2 versus L1 reading performance; these reductions will also include larger L2 word frequency effects. Third, both groups of children will exhibit reduced global and local reading performance compared with adults (across their known languages); these reductions will also include larger word frequency effects. Across all three predictions, reductions will be driven by lower-frequency words and reflected in both fixation and saccadic measures (e.g., longer fixation durations, more saccades).

1

However, this prediction has never been explicitly made nor tested by the model’s authors.

322

V. Whitford, M.F. Joanisse / Journal of Experimental Child Psychology 173 (2018) 318–337

Method Participants In total, 34 English monolingual children and 33 English–French bilingual children aged 7 to 12 years participated in the study. Ethics approval was obtained from the University of Western Ontario’s nonmedical research ethics board. Participants were recruited from elementary schools in the London area of Ontario, Canada, and received movie gift cards as compensation. Monolingual children attended English schools and had minimal functional knowledge and use of an L2 (French). Note that basic French is taught in English-language elementary schools in Ontario beginning in Grade 4 (10 years of age). Thus, older monolingual children had some exposure to French, albeit at a very low level that would not meet the standard of a bilingual or L2 speaker. Bilingual children attended either French or French immersion schools and were proficient in both their L1 (English) and L2 (French). Participants’ demographic and language backgrounds were assessed using a parental adaptation of the Language Experience and Proficiency Questionnaire (LEAP-Q; Marian, Blumenfeld, & Kaushanskaya, 2007). Participants’ word-level reading skills were assessed using the Word Reading and Pseudoword Decoding subtests of the Wechsler Individual Achievement Test–Second Edition (WIAT-II, English–Canadian and French–Canadian adaptations; Wechsler, 2005). Participants’ nonverbal IQ was assessed using the Test of Nonverbal Intelligence–Third Edition (TONI-III; Brown, Sherbenou, & Johnsen, 1997). Participant characteristics are presented in Table 1, which demonstrates that the two groups of children were matched on age, education, sex, parental socioeconomic status (SES) based on the Hollingshead Occupational Scale (Hollingshead, 1975), and nonverbal IQ (all p values > .05). Nonverbal IQ was within the normal range (i.e., 100 ± 15). The two groups of children were also matched on parental-report (LEAP-Q) and objective (WIAT-II) measures of L1 history and reading ability (all p values > .05). Objective L1 reading ability was within the normal range (i.e., 100 ± 15). Expectedly, however, they significantly differed on parental-report measures of L2 history and reading ability (all p values < .001); monolingual children lacked the proficiency needed to complete the objective measures of L2 reading ability. Bilingual children had significantly lower L2 versus L1 reading ability based on both parental-report and objective measures (all p values < .05). For comparison purposes, 30 English monolingual adults and 30 English–French bilingual adults aged 18 to 21 years also participated in the study. Participants were recruited from the University of Western Ontario, and received course credit as compensation. Monolingual adults had minimal functional knowledge and use of an L2 (French); bilingual adults were proficient in both their L1 (English) and L2 (French). Participant characteristics are presented in Table 2, which demonstrates that the two groups of adults were matched on age, education, sex, parental SES, and nonverbal IQ (all p values > .05). They were also matched on self-report (LEAP-Q) and objective (WIAT-II) measures of L1 history and reading ability (all p values > .05). However, they significantly differed on self-report measures of L2 history and reading ability (all p values < .001). Bilingual adults had significantly lower L2 versus L1 reading ability based on both self-report and objective measures (all p values < .001). All participants included in the study were typically developing, with no uncorrected visual or hearing impairments and no language, learning, neurological, or psychiatric disorders. The groups of monolingual children and adults were matched as closely as possible, including on parental SES (3.00 ± 1.18 vs. 2.63 ± 1.14; p > .05) and Pseudoword Decoding (106.26 ± 15.61 vs. 105.73 ± 11.85; p > .05). However, they significantly differed on Word Reading (99.44 ± 12.58 vs. 111.80 ± 6.55; p < .001) and nonverbal IQ (109.88 ± 17.10 vs. 99.60 ± 11.84; p < .01). Similarly, the groups of bilingual children and adults were also matched as closely as possible, including on parental SES (2.88 ± 1.36 vs. 2.27 ± 1.09; p > .05), L1 Pseudoword Decoding (103.12 ± 17.22 vs. 109.07 ± 8.12; p > .05), L2 Word Reading (88.55 ± 23.77 vs. 81.18 ± 18.92; p > .05), and L2 Pseudoword Decoding (95.70 ± 20.73 vs. 97.70 ± 12.78; p > .05). However, like the monolingual age groups, they also significantly differed on L1 Word Reading (99.15 ± 17.38 vs. 112.43 ± 5.39; p < .001) and nonverbal IQ (117.18 ± 18.04 vs. 99.39 ± 14.04; p < .001). The age group differences in L1 word reading skills

V. Whitford, M.F. Joanisse / Journal of Experimental Child Psychology 173 (2018) 318–337

323

Table 1 Child participant characteristics. Monolingual children (n = 34) [mean (SD)]

Bilingual children (n = 33) [mean (SD)]

Age (years) Sex (male:female) Education (years) Parental SESa

9.82 (1.10) 14:20 4.09 (1.08) 3.00 (1.18)

10.02 (1.32) 13:20 4.21 (1.39) 2.88 (1.36)

AoA; Age of fluency (years) L1 L2***

Birth (–); 2.71 (0.95) 7.42 (1.82); Never (–)

Birth (–); 2.43 (1.17) 3.82 (1.66); 5.57 (1.96)

Reading AoA; Age of reading fluency (years) L1 4.35 (0.96); 6.05 (0.95) L2*** 8.28 (1.07); Never (–)

4.48 (1.14); 6.23 (1.37) 5.47 (1.05); 7.36 (1.44)

Current language exposure (% time) L1*** 95.53 (5.66) L2*** 4.47 (5.66)

58.03 (12.93) 39.70 (13.11)

Current reading exposure (% time) 99.79 (0.88) L1*** L2*** 0.21 (0.88)

65.30 (25.98) 33.58 (25.35)

L1 self-report proficiency measures (1–7)b Reading ability 6.06 (1.41) Overall competence 6.15 (1.31)

5.64 (1.39) 5.88 (1.11)

L2 self-report proficiency measures (1–7)b Reading ability*** 1.06 (0.24) Overall competence*** 1.06 (0.24)

4.58 (1.28) 4.67 (1.31)

L1 WIAT-II (standard scores) Word Reading Pseudoword Decoding

99.44 (12.58) 106.26 (15.61)

99.15 (17.38) 103.12 (17.22)

L2 WIAT-II (standard scores) Word Reading Pseudoword Decoding

– –

88.55 (23.77) 95.70 (20.73)

TONI-III (standard scores)

109.88 (17.10)

117.18 (18.04)

Note. SES, socioeconomic status; AoA, age of acquisition; L1, first-language; L2, second-language; WIAT-II, Wechsler Individual Achievement Test–Second Edition; TONI-III, Test of Nonverbal Intelligence–Third Edition. a Scale from 1 (major professional) to 9 (unemployed). b Scale from 1 (beginner) to 7 (native-like). *** p < .001.

suggest that our samples of adults might represent highly skilled expert readers (as compared with the general population). Moreover, the age group differences in nonverbal IQ suggest that our samples of children might have been selected from the slightly higher end of the distribution.

Materials Stimuli were four paragraphs reflective of the type of reading conducted in general education classrooms across Canada. The paragraphs were English and French versions of fiction and nonfiction articles taken from the Reading Comprehension subtest of the WIAT-II (Wechsler, 2005). The fiction articles were about a character who loved cleaning their town and a character who raised money for hospitalized children; the nonfiction articles were about crickets and baobab trees. As such, the paragraphs were ecologically relevant texts. The four English paragraphs contained 105, 87, 103, and 195 words, and the four French paragraphs contained 118, 95, 109, and 200 words. The words of each paragraph were coded for length, frequency, and predictability (see Table A1 of the Appendix for paragraph characteristics). English word frequencies (subtitle values in occurrences per million words) were obtained from the Brysbaert and New

324

V. Whitford, M.F. Joanisse / Journal of Experimental Child Psychology 173 (2018) 318–337

Table 2 Adult participant characteristics. Monolingual adults (n = 30) [mean (SD)]

Bilingual adults (n = 30) [mean (SD)]

Age (years) Sex (male:female) Education (years) Parental SESa

18.67 (0.94) 10:20 13.35 (0.52) 2.63 (1.14)

18.33 (0.60) 5:25 13.17 (0.30) 2.27 (1.09)

AoA; Age of fluency (years) L1 L2***

Birth (–); 3.72 (1.80) 8.96 (2.46); Never (–)

Birth (–); 3.68 (1.75) 5.53 (2.42); 10.95 (4.51)

Reading AoA; Age of reading fluency (years) L1 4.52 (1.32); 6.50 (1.96) L2*** 9.90 (2.24); Never (–)

5.03 (1.49); 7.07 (1.57) 7.13 (2.08); 11.10 (3.56)

Current language exposure (% time) L1** 99.70 (0.97) L2*** 0.30 (0.97)

86.41 (19.71) 12.73 (19.67)

Current reading exposure (% time) 100.00 (0.00) L1*** L2*** 0.00 (0.00)

86.34 (18.84) 13.66 (18.84)

L1 self-report proficiency measures (1–7)b Reading ability 6.83 (0.73) Overall competence 6.93 (0.36)

6.67 (0.74) 6.70 (0.53)

L2 self-report proficiency measures (1–7)b Reading ability*** 1.43 (0.62) Overall competence*** 1.17 (0.37)

5.20 (0.98) 4.83 (1.07)

L1 WIAT-II (standard scores) Word Reading Pseudoword Decoding

111.80 (6.55) 105.73 (11.85)

112.43 (5.39) 109.07 (8.12)

L2 WIAT-II (standard scores) Word Reading Pseudoword Decoding

– –

81.18 (18.92) 97.70 (12.78)

TONI-III (standard scores)

99.60 (11.84)

99.39 (14.04)

Note. SES, socioeconomic status; AoA, age of acquisition; L1, first-language; L2, second-language; WIAT-II, Wechsler Individual Achievement Test–Second Edition; TONI-III, Test of Nonverbal Intelligence–Third Edition. a Scale from 1 (major professional) to 9 (unemployed). b Scale from 1 (beginner) to 7 (native-like). ** p < .01. *** p < .001.

(2009) corpus within the English Lexicon Project (Balota et al., 2007); French word frequencies (subtitle values in occurrences per million words) were obtained from the Lexique database (New, Pallier, Ferrand, & Matos, 2001). English and French word predictabilities (cloze values) were obtained through computerized tasks in which separate samples of native English (n = 30) and native French (n = 30) speakers guessed the words of each paragraph (in a sequential manner and one word at time) until each paragraph was presented in its entirety on the screen (following Miellet, Sparrow, & Sereno, 2007; Whitford & Titone, 2012, 2014, 2017b). Correct guesses were coded as 1, and incorrect guesses were coded as 0. Average cloze values were then computed for each word. A total of 210 target words were selected from the four paragraphs for the local (i.e., word-level) analyses. This total excluded line-initial and line-final words; function, punctuated, and repeated words; proper nouns; and words with cross-language orthographic overlap, such as cognates and interlingual homographs (following Miellet et al., 2007; Pollatsek et al., 2006; Whitford & Titone, 2012, 2014, 2017b). As such, the target words consisted of language-specific content words (see Table A1 of the Appendix for target word characteristics).

V. Whitford, M.F. Joanisse / Journal of Experimental Child Psychology 173 (2018) 318–337

325

Apparatus An EyeLink 1000 desktop-mounted eye-tracker (SR Research, Ottawa, Ontario, Canada) recorded eye movements at a 1-kHz sampling rate. Viewing was binocular; however, tracking was right-eye monocular. The paragraphs were displayed on a 21-inch ViewSonic CRT monitor, with a 1024  768 pixel screen resolution, placed 60 cm in front of participants. The paragraphs were double-spaced and presented on either one or two display screens (i.e., ‘‘pages”) in yellow 14-point Courier New font against a black background (Experiment Builder, SR Research). The display screens had a maximum of 10 lines of text, 70 characters per line, and 2 characters per 1° of visual angle. Calibration was performed with a 9-point grid; the average fixation error was <0.5° of visual angle. A padded headrest minimized head movements during reading. Procedure Participants first completed the eye-tracking reading task in which they read the four paragraphs silently and naturally for comprehension. Monolingual participants read the four paragraph versions in English; bilingual participants read two paragraph versions in English (their L1) and two in French (their L2). Paragraph version (1, 2, 3, 4) was counterbalanced across participants using a Latin square design; paragraph language (L1, L2) was also randomly counterbalanced across bilingual participants. Comprehension was measured through orally administered, open-ended questions (four per paragraph, designed by the experimenters). Correct, partially correct, and incorrect answers were coded as 1, 0.5, and 0, respectively (following Radach, Huestegge, & Reilly, 2008; Whitford & Titone, 2012, 2014, 2017b). Participants then completed the LEAP-Q, followed by the WIAT-II Word Reading and Pseudoword Decoding subtests, and finally the TONI-III. Results Comprehension performance We ran t tests which revealed comparable L1 accuracy in the monolingual and bilingual children (84% vs. 82%, p = .38) and comparable L1 and L2 accuracy in the bilingual children (82% vs. 78%, p = .40). Moreover, the children’s accuracy was comparable to that of their respective adult comparison groups across the L1 (monolinguals: 84% vs. 86%, p = .44; bilinguals: 82% vs. 87%, p = .17) and L2 (bilinguals: 78% vs. 81%, p = .55). Eye movement data The EyeLink 1000 software uses a combined acceleration and velocity algorithm to parse out fixations (pauses) and saccades (eye movements); saccades had a minimum velocity of 30°/s, minimum acceleration of 8000°/s2, and minimum change in eye position of 0.15°. An 80-ms lower cutoff was applied to all fixations (<5% of all data). An upper cutoff was not applied to maximize data inclusion; the maximum fixation duration was 2605 ms (which came from a bilingual child reading a lowerfrequency L2 word). Both global (i.e., text-level) and local (i.e., word-level) eye movement measures were examined (reviewed in Rayner, 1998, 2009; Rayner et al., 2012; Whitford et al., 2016). Our global measures, which reflect processing difficulty across all words in all four paragraphs, included reading rate (number of words read per minute), average fixation duration (in ms), total number of saccades (both progressive and regressive), total number of words fixated (an index of skipping rate), and total reading time (in ms). Our local measures, which reflect processing difficulty for the 210 target words selected from the four paragraphs, included first fixation duration (i.e., duration of the very first fixation on a word), gaze duration (i.e., sum of all fixation durations on a word during the first pass), skipping rate (i.e., probability of fixating a word during the first pass), regressions out (i.e., probability of regressing out of a word

326

V. Whitford, M.F. Joanisse / Journal of Experimental Child Psychology 173 (2018) 318–337

to an earlier occurring word), and total reading time (i.e., sum of all fixation and refixation durations on a word). First fixation duration, gaze duration, and skipping rate are considered early-stage local measures that reflect lexical access; regressions out and total reading time are considered late-stage local measures that reflect postlexical integration. These measures were selected as those typically reported in bilingual eye movement reading studies (e.g., Cop, Drieghe et al., 2015; Cop, Keuleers et al., 2015; Gollan et al., 2011; Whitford & Titone, 2012, 2015, 2016, 2017a,b). Global (i.e., text-level) analyses We conducted four analyses. The first two examined global aspects of L1 reading performance in (a) monolingual versus bilingual children and (b) children versus adults. The second two examined global aspects of L2 reading performance in (a) bilingual children (relative to their L1) and (b) bilingual children versus adults. The data were analyzed using linear mixed-effects models (LMMs) within the lme4 package (Bates, Mächler, Bolker, & Walker, 2014) of R (Version 3.4.1) (Baayen, 2008; Baayen, Davidson, & Bates, 2008; R Development Core Team., 2017). The specifications of each model, such as fixed effects (i.e., factors of theoretical interest), control predictors (i.e., covariates), and random effects (i.e., random intercepts and/or slopes for participants and items), are reported for each analysis to follow. The same model was applied to all eye movement measures within each analysis. Across all models, each categorical variable was deviation coded ( 0.5, 0.5) (i.e., the mean of each level was compared with the grand mean), and each continuous variable was scaled (i.e., standardized, z-scored) to reduce collinearity. Significant effects of interest (reported subsequently) are those with |t| values > 1.96, corresponding to a = .05. To facilitate comparison with traditional analyses of variance (ANOVAs), participant means (for categorical variables) are presented in conjunction with significant effects. L1 reading performance in monolingual and bilingual children The fixed factor was language group (monolingual children vs. bilingual children). Control predictors included age (continuous), L1 Word Reading (continuous), total number of words (per paragraph), average word length (per paragraph), average word frequency (per paragraph), and average word predictability (per paragraph). Of note, age and L1 Word Reading were included to control for any variance due to the wide range in age and, consequently, variability in reading experience. Random factors included random intercepts for participants and paragraph version (Barr, Levy, Scheepers, & Tily, 2013; Bates, Kliegl, Vasishth, & Baayen, 2015). Sample syntax for this analysis is provided subsequently: Sample Model = lmer (eye movement measure  language group + age + L1 Word Reading + number of words + word length + word frequency + word predictability + (1 | participant) + (1 | paragraph version)) A significant effect of language group was found for all eye movement measures except average fixation duration. Specifically, monolingual children had faster reading rates (183 vs. 142 words/min; b = 40.71, SE = 13.48, t = 3.02), fewer saccades (135 vs. 159; b = 24.33, SE = 8.47, t = 2.87), shorter total reading times (37,643 vs. 46,831 ms; b = 9257.92, SE = 3201.31, t = 2.89), and fixated fewer words (77 vs. 84; b = 5.67, SE = 2.32, t = 2.45) than bilingual children during L1 reading. Of note, all language group differences persist even when restricting the analyses to monolingual children with absolutely no L2 experience. Complete model outputs can be found in Table A2 of the Appendix. L1 reading performance in children and adults The fixed factor was age group (monolingual children vs. monolingual adults or bilingual children vs. bilingual adults). Control predictors included L1 Word Reading (continuous), nonverbal IQ (continuous), total number of words (per paragraph), average word length (per paragraph), average word frequency (per paragraph), and average word predictability (per paragraph). Of note, L1 Word Reading and nonverbal IQ were included to control for the age group differences on these measures. Random

V. Whitford, M.F. Joanisse / Journal of Experimental Child Psychology 173 (2018) 318–337

327

factors included random intercepts for participants and paragraph version. Sample syntax for this analysis is provided subsequently: Sample Model = lmer (eye movement measure  age group + L1 Word Reading + nonverbal IQ + number of words + word length + word frequency + word predictability + (1 | participant) + (1 | paragraph version)) A significant effect of age group was found for all eye movement measures except number of words fixated. Specifically, both groups of children had slower reading rates (monolinguals: 183 vs. 204 words/min; b = 220.89, SE = 70.91, t = 3.12; bilinguals: 142 vs. 210 words/min; b = 51.26, SE = 19.02, t = 2.70), longer fixation durations (monolinguals: 239 vs. 202 ms; b = 26.96, SE = 7.82, t = 3.45; bilinguals: 256 vs. 211 ms; b = 26.59, SE = 11.87, t = 2.24), and more saccades (monolinguals: 135 vs. 129; b = 118.33, SE = 41.25, t = 2.87; bilinguals [marginal]: 159 vs. 122; b = 23.24, SE = 12.08, t = 1.92) than adults during L1 reading. Bilingual children also had longer total reading times (46,831 vs. 30,744 ms; b = 10,603.37, SE = 4499.03, t = 2.36) than adults. Complete model outputs can be found in Tables A3 and A4 of the Appendix. L1 and L2 reading performance in bilingual children The fixed factor was paragraph language (L1 vs. L2). Control predictors included age (continuous), L1 Word Reading (continuous), L2 Word Reading (continuous), paragraph language version (English vs. French), total number of words (per paragraph), average word length (per paragraph), average word frequency (per paragraph), and average word predictability (per paragraph). Of note, L2 Word Reading and paragraph language version were included to control for any variance due to differences in L2 experience and linguistic differences between English and French, respectively. Random factors included random intercepts for participants and paragraph version and random slopes for paragraph language across participants and paragraph version. Sample syntax for this analysis is provided subsequently: Sample Model = lmer (eye movement measure  paragraph language + age + L1 Word Reading + L2 Word Reading + paragraph language version + number of words + word length + word frequency + word predictability + (1 + paragraph language | participant) + (1 + paragraph language | paragraph version)) A significant effect of paragraph language was found for all eye movement measures except reading rate and average fixation duration. Specifically, bilingual children had more saccades (194 vs. 159; b = 25.91, SE = 11.37, t = 2.28), longer total reading times (59,072 vs. 46,831 ms; b = 9299.01, SE = 3810.88, t = 2.44), and fixated more words (89 vs. 84; b = 4.79, SE = 1.71, t = 2.79) during L2 versus L1 reading. Complete model outputs can be found in Table A5 of the Appendix. L2 reading performance in bilingual children and adults The fixed factor was age group (bilingual children vs. bilingual adults). Control predictors included L2 Word Reading (continuous), nonverbal IQ (continuous), total number of words (per paragraph), average word length (per paragraph), average word frequency (per paragraph), and average word predictability (per paragraph). Random factors included random intercepts for participants and paragraph version. Sample syntax for this analysis is provided subsequently: Sample Model = lmer (eye movement measure  age group + L2 Word Reading + nonverbal IQ + number of words + word length + word frequency + word predictability + (1 | participant) + (1 | paragraph version)) A significant effect of age group was found for all eye movement measures except number of saccades and number of words fixated. Specifically, bilingual children had slower reading rates (126 vs. 142 words/min; b = 25.64, SE = 12.32, t = 2.08), longer fixation durations (260 vs. 233 ms; b = 29.49, SE = 9.75, t = 3.03), and longer total reading times (59,072 vs. 49,395 ms; b = 12,732.06, SE = 6118.63, t = 2.08) than adults during L2 reading. Complete model outputs can be found in Table A6 of the Appendix.

328

V. Whitford, M.F. Joanisse / Journal of Experimental Child Psychology 173 (2018) 318–337

Summary of global (i.e., text-level) analyses We found that bilingual children exhibit reduced L1 reading performance relative to monolingual children, bilingual children exhibit reduced L2 versus L1 reading performance, and both groups of children exhibit reduced reading performance relative to adults (across their L1 and L2). Local (i.e., word-level) analyses: Examination of word frequency effects We conducted four analyses. The first two examined local aspects of L1 reading performance in (a) monolingual versus bilingual children and (b) children versus adults. The second two examined local aspects of L2 reading performance in (a) bilingual children (relative to their L1) and (b) bilingual children versus adults. The data were analyzed using LMMs; however, logistic generalized linear mixed-effects models (GLMMs) were used for skipping rate and regressions out given their binomial nature. Following the global (i.e., text-level) analyses, the same model was applied to all eye movement measures within each analysis (specified subsequently), each categorical variable was deviation coded ( 0.5, 0.5), and each continuous variable was scaled. Significant effects of interest (reported subsequently) are those with |t|- or |z|-score values > 1.96 for LMMs and logistic GLMMs, respectively, and are accompanied by participant means (for categorical variables). L1 reading performance in monolingual and bilingual children Fixed factors included language group (monolingual children vs. bilingual children) and word frequency (continuous and log-transformed). Control predictors included age (continuous), L1 Word Reading (continuous), word length (continuous), and word predictability (continuous). Random factors included random intercepts for participants and paragraph version. Sample syntax for this analysis is provided subsequently: Sample Model = lmer (eye movement measure  language group ⁄ word frequency + age + L1 Word Reading + word length + word predictability + (1 | participant) + (1 | paragraph version)) A significant effect of language group was found for all eye movement measures except skipping rate and regressions out. Specifically, monolingual children had shorter first fixation durations (243 vs. 271 ms; b = 27.92, SE = 8.93, t = 3.13), gaze durations (325 vs. 377 ms; b = 51.49, SE = 17.93, t = 2.87), and total reading times (435 vs. 541 ms; b = 105.71, SE = 30.23, t = 3.50) than bilingual children during L1 reading. Of note, all language group differences persist even when restricting the analyses to monolingual children with absolutely no L2 experience. A significant effect of word frequency was also found for all eye movement measures except regressions out, which was marginal. Specifically, as word frequency increased, skipping rates increased (b = 0.19, SE = 0.07, z = 2.80), whereas first fixation durations (b = 18.61, SE = 1.96, t = 9.48), gaze durations (b = 41.82, SE = 3.41, t = 12.28), regressions out (b = 0.08, SE = 0.04, z = 1.95), and total reading times (b = 67.37, SE = 4.90, t = 13.74) decreased. Lastly, a significant interaction between language group and word frequency was found for all eye movement measures except skipping rate and regressions out. As can be seen in Fig. 1, as word frequency increased, both groups of children showed decreases in first fixation durations (b = 12.53, SE = 3.57, t = 3.51), gaze durations (b = 17.35, SE = 6.18, t = 2.81), and total reading times (b = 33.36, SE = 8.99, t = 3.71); however, the processing difference between lower-frequency and higherfrequency words was larger for bilingual children, reflecting a larger L1 word frequency effect. More specifically, bilingual children were differentially slower at processing lower-frequency L1 words. Complete model outputs can be found in Table A7 of the Appendix. L1 reading performance in children and adults Fixed factors included age group (monolingual children vs. monolingual adults or bilingual children vs. bilingual adults) and word frequency (continuous and log-transformed). Control predictors included L1 Word Reading (continuous), nonverbal IQ (continuous), word length (continuous), and word predictability (continuous). Random factors included random intercepts for participants and paragraph version. Sample syntax for this analysis is provided subsequently:

V. Whitford, M.F. Joanisse / Journal of Experimental Child Psychology 173 (2018) 318–337

329

Fig. 1. The effect of word frequency on monolingual and bilingual children’s first fixation durations (top left panel), gaze durations (top right panel), and total reading times (bottom panel) during L1 reading. Actual values are plotted. Shaded areas represent confidence intervals.

Sample Model = lmer (eye movement measure  age group ⁄ word frequency + L1 Word Reading + nonverbal IQ + word length + word predictability + (1 | participant) + (1 | paragraph version)) A significant effect of age group was found for first fixation duration and gaze duration, where both groups of children had longer first fixation durations (monolinguals: 243 vs. 210 ms; b = 23.64, SE = 8.38, t = 2.82; bilinguals: 271 vs. 216 ms; b = 38.74, SE = 13.19, t = 2.94) and gaze durations (monolinguals: 325 vs. 250 ms; b = 51.42, SE = 17.56, t = 2.93; bilinguals: 377 vs. 253 ms; b = 74.13, SE = 25.44, t = 2.91) than adults during L1 reading. A significant effect of word frequency was also found for all eye movement measures except regressions out; the details of this effect are not repeated for parsimony (see Tables A8 and A9 of the Appendix for complete model outputs). Lastly, a significant interaction between age group and word frequency was found for first fixation duration (bilinguals only: b = 15.09, SE = 4.28, t = 3.53), gaze duration (monolinguals: b = 21.41, SE = 4.04, t = 5.30; bilinguals: b = 46.26, SE = 7.09, t = 6.53), and total reading time (monolinguals: b = 38.49, SE = 6.29, t = 6.12; bilinguals: b = 70.69, SE = 11.50, t = 6.15). As can be seen in Figs. 2 and 3, both groups of children showed larger word frequency effects than adults, which were driven by their differentially slower processing of lower-frequency L1 words. L1 and L2 reading performance in bilingual children Fixed factors included paragraph language (L1 vs. L2) and word frequency (continuous and logtransformed). Control predictors included age (continuous), L1 Word Reading (continuous), L2 Word Reading (continuous), paragraph language version (English vs. French), word length (continuous), and word predictability (continuous). Random factors included random intercepts for participants and paragraph version and random slopes for paragraph language across participants and paragraph version. Sample syntax for this analysis is provided subsequently:

330

V. Whitford, M.F. Joanisse / Journal of Experimental Child Psychology 173 (2018) 318–337

Fig. 2. The effect of word frequency on monolingual children’s and adults’ gaze durations (left panel) and total reading times (right panel) during L1 reading. Actual values are plotted. Shaded areas represent confidence intervals.

Fig. 3. The effect of word frequency on bilingual children’s and adults’ first fixation durations (top left panel), gaze durations (top right panel), and total reading times (bottom panel) during L1 reading. Actual values are plotted. Shaded areas represent confidence intervals.

Sample Model = lmer (eye movement measure  paragraph language ⁄ word frequency + age + L1 Word Reading + L2 Word Reading + paragraph language version + word length + word predictability + (1 + paragraph language | participant) + (1 + paragraph language | paragraph version)) A significant effect of paragraph language was found for regressions out and total reading time, where bilingual children had more regressions out (0.29 vs. 0.26; b = 0.33, SE = 0.16, z = 2.09) and longer total reading times (687 vs. 541 ms; b = 99.42, SE = 35.68, t = 2.79) during L2 versus L1 reading. A significant effect of word frequency was also found for all eye movement measures except skipping rate and regressions out; the details of this effect are not repeated for parsimony (see Table A10 of the Appendix for complete model outputs). Lastly, a significant interaction between paragraph language and word frequency was found for gaze duration and total reading time. As can be seen in Fig. 4, as word frequency increased, gaze durations (b = 25.26, SE = 11.00, t = 2.30) and total reading times (b = 31.22, SE = 15.44, t = 2.02)

V. Whitford, M.F. Joanisse / Journal of Experimental Child Psychology 173 (2018) 318–337

331

Fig. 4. The effect of word frequency on bilingual children’s gaze durations (left panel) and total reading times (right panel) during L1 and L2 reading. Actual values are plotted. Shaded areas represent confidence intervals.

decreased in both languages; however, the processing difference between lower-frequency and higher-frequency words was larger for L2 reading, reflecting a larger L2 word frequency effect. More specifically, bilingual children were differentially slower at processing lower-frequency L2 words. L2 reading performance in bilingual children and adults Fixed factors included age group (bilingual children vs. bilingual adults) and word frequency (continuous and log-transformed). Control predictors included L2 Word Reading (continuous), nonverbal IQ (continuous), word length (continuous), and word predictability (continuous). Random factors included random intercepts for participants and paragraph version. Sample syntax for this analysis is provided subsequently: Sample Model = lmer (eye movement measure  age group ⁄ word frequency + L2 Word Reading + nonverbal IQ + word length + word predictability + (1 | participant) + (1 | paragraph version)) A significant effect of age group was found for first fixation duration and gaze duration, where bilingual children had longer first fixation durations (271 vs. 249 ms; b = 26.51, SE = 12.51, t = 2.12) and gaze durations (446 vs. 364 ms; b = 73.65, SE = 34.52, t = 2.13) than adults during L2 reading. A significant effect of word frequency was also found for all eye movement measures except regressions out; the details of this effect are not repeated for parsimony (see Table A11 of the Appendix for complete model outputs). Lastly, a significant interaction between age group and word frequency was found for first fixation duration (b = 19.99, SE = 4.78, t = 4.18) and gaze duration (b = 40.05, SE = 10.86, t = 3.69). As can be seen in Fig. 5, bilingual children showed larger word frequency effects than adults, which were driven by their differentially slower processing of lower-frequency L2 words. Summary of local (i.e., word-level) analyses We found that bilingual children exhibit reduced L1 reading performance relative to monolingual children, including larger L1 word frequency effects; bilingual children exhibit reduced L2 versus L1 reading performance, including larger L2 word frequency effects; and both groups of children exhibit reduced reading performance relative to adults (across their L1 and L2), including larger word frequency effects. Discussion There has been a recent upsurge in research examining children’s eye movements to more accurately capture naturalistic reading situations (reviewed in Blythe & Joseph, 2011; Frey, 2016; Rayner, 1998, 2009; Rayner et al., 2012; Reichle et al., 2013), providing useful insights into online reading processes that are not identifiable using more traditional measures, such as standardized tests and response-based tasks. Indeed, eye-tracking research has furthered our understanding of the per-

332

V. Whitford, M.F. Joanisse / Journal of Experimental Child Psychology 173 (2018) 318–337

Fig. 5. The effect of word frequency on bilingual children’s and adults’ first fixation durations (left panel) and gaze durations (right panel) during L2 reading. Actual values are plotted. Shaded areas represent confidence intervals.

ceptual, oculomotor, cognitive, and linguistic processes underlying reading acquisition. However, work to date has almost entirely focused on reading in monolingual children. Thus, the extent to which these findings extend to bilingual children is largely unknown. An important consequence of bilingual children acquiring two reading systems simultaneously is that they necessarily have less experience and practice with each reading system than their monolingual peers. This divided experience and practice, in turn, may lead to differences in their eye movement record, ultimately reflecting differences in how they process individual words and connected texts. The current study investigated this issue; it examined eye movement measures of global (i.e., text-level) and local (i.e., word-level) reading performance in monolingual and bilingual children (across their known languages) and included samples of monolingual and bilingual young adults to allow for developmental comparisons. The current study also focused on differences in word frequency effects because they provide an index of lexical accessibility, that is, they reflect how easily words are accessed from memory. As discussed below, we identified three main findings pertaining to (a) differences between monolingual and bilingual children during L1 reading, (b) differences between L1 and L2 reading in bilingual children, and (c) differences between children and adults.

Differences between monolingual and bilingual children in L1 reading First, we found that both global and local measures of L1 reading performance were reduced in bilingual versus monolingual children. Global reductions included slower reading rates, more saccades (both progressive and regressive), longer total reading times, and more fixated words; local reductions, which spanned both early and late stages of reading, included longer first fixation durations, gaze durations, and total reading times. Also with respect to local processing, bilingual children displayed larger L1 word frequency effects during both early-stage and late-stage reading. This finding was driven by their differentially slower processing of lower-frequency words (i.e., words that are less lexically entrenched). Corroborating the weaker links hypothesis (Gollan et al., 2008, 2011), these findings suggest that bilingual children’s lexical representations for lower-frequency L1 words, which have benefited from less absolute exposure than those of monolingual children (and, thus, are further from a functional ceiling), experience reduced quality and accessibility (i.e., weaker links), resulting in larger L1 word frequency effects. In contrast, higher-frequency L1 words were comparably processed by the two groups, reflecting similar levels of exposure to such words. These reductions in word-level processing, in turn, might have scaled up to affect text-level measures of reading performance, ultimately resulting in more effortful L1 reading. We note, however, that bilingual children’s L1 reading performance might also have been influenced by other factors not directly examined in the current study, including relatively greater difficulty with L1 syntactic processing. Relatedly, bilingual children’s reduced L1 reading performance might also reflect the more limited linguistic environments in which they encounter print. Unlike their monolingual peers, who may encounter print across a variety of diverse linguistic environments (e.g., at home, at school, in the

V. Whitford, M.F. Joanisse / Journal of Experimental Child Psychology 173 (2018) 318–337

333

community), bilingual children may encounter print in a more compartmentalized and/or restricted manner (e.g., L1 at home only, L2 at school only), resulting in less semantic diversity overall and, consequently, less entrenched word-related knowledge. This view fits well with the lexical legacy hypothesis (Nation, 2017), which was implicitly developed for monolingual children but can be extended to bilingual children. These findings are also consistent with recent (monolingual) modeling work demonstrating that children’s eye movements during reading are driven by their rates of lexical processing (see Mancheva et al., 2015; Reichle et al., 2013) and extend recent (monolingual) experimental work demonstrating that children are sensitive to the effects of word frequency during natural reading (reviewed in Blythe & Joseph, 2011; Frey, 2016; Rayner, 1998, 2009; Rayner et al., 2012; Reichle et al., 2013; cf. Blythe et al., 2006). These findings also shed light on how different patterns of word frequency effects (and reading more generally) can emerge between monolinguals and bilinguals during L1 reading. For instance, prior eye movement research has reported comparable L1 word frequency effects (Cop, Keuleers et al., 2015) and sentence-level reading behavior (Cop, Drieghe et al., 2015) in monolingual and bilingual young adults. It was proposed that comparable levels of L1 proficiency, and, by extension, comparable levels of L1 exposure, can lead to similar L1 word frequency effects between the two groups. Here, our groups of monolingual and bilingual children also had comparable levels of L1 proficiency based on both parental-report (LEAP-Q) and objective (WIAT-II Word Reading and Pseudoword Decoding subtests) measures. However, despite this, they significantly differed in terms of current L1 exposure (99.79% vs. 65.30%, p < .001). Thus, weaker links and, consequently, word frequency effects can indeed vary between monolinguals and bilinguals as a function of their current amounts of L1 experience. Importantly, these findings are inconsistent with an extensive body of work reporting largely comparable (and at times better) word-level reading skills in bilingual versus monolingual children, particularly when bilinguals acquire their L2 early in life and/or during later childhood years (e.g., August & Shanahan, 2006; Berens, Kovelman, & Petitto, 2013; Jared, 2015). Text-level reading skills, in contrast, have been inconsistently reported across studies (e.g., August & Shanahan, 2006; Jared, 2015). We emphasize, however, that prior work has primarily focused on offline reading measures, such as standardized tests, which may lack the sensitivity to detect persistent differences in how monolingual and bilingual children mentally represent words and read connected texts. Indeed, our groups of monolingual and bilingual children performed comparably on our offline reading measures (i.e., WIAT-II Word Reading and Pseudoword Decoding subtests), which signals the importance of using more naturalistic and temporally sensitive measures, such as eye movement recordings, when assessing reading in these populations. Lastly, we highlight here that the ultimate goal of reading is to extract meaning from print, that is, to comprehend what is read. Despite their differences in the eye movement record, monolingual and bilingual children had comparable levels of L1 reading comprehension (for the experimental paragraphs). Thus, one possibility is that bilingual children allocate more resources during L1 reading to compensate for their divided language experience (e.g., by devoting more time at each fixation) and, ultimately, achieve similar levels of comprehension as their monolingual peers. Accordingly, between-group differences in the eye movement record might not necessarily reflect deficits but rather different (and potentially adaptive) reading strategies (see also Whitford & Titone, 2017b, who provide a similar account for age-related differences in word-level processing between bilingual younger and older adults).

Differences between L1 and L2 reading in bilingual children Second, we found that both global and local measures of L2 versus L1 reading performance were reduced in bilingual children. Global reductions included more saccades (both progressive and regressive), longer total reading times, and more fixated words; local reductions included more regressions out and longer total reading times, as well as larger word frequency effects during both early-stage and late-stage reading. This latter finding was driven by their differentially slower processing of lower-frequency L2 words (i.e., words that are, by far, less lexically entrenched).

334

V. Whitford, M.F. Joanisse / Journal of Experimental Child Psychology 173 (2018) 318–337

Corroborating the weaker links hypothesis (Gollan et al., 2008, 2011), these findings suggest that bilingual children’s lexical representations for lower-frequency L2 words, which have benefited from far less exposure than those of L1 words (and, thus, are even further from a functional ceiling), experience reduced quality and accessibility (i.e., weaker links), resulting in larger L2 word frequency effects. In contrast, higher-frequency words were comparably processed in both languages, reflecting similar levels of exposure to such words. These reductions in word-level processing, in turn, might have scaled up to affect text-level measures of reading performance, ultimately resulting in more effortful L2 reading. We note, however, that bilingual children’s L2 reading performance might also have been influenced by other factors not directly examined in the current study, including relatively greater difficulty with L2 syntactic processing. Bilingual children’s reduced L2 reading performance might also reflect the more limited linguistic contexts in which they encounter L2 print, resulting in less entrenched L2 word-related knowledge (Nation, 2017). These findings are also consistent with recent eye movement work reporting larger L2 versus L1 word frequency effects in bilingual younger (Cop, Keuleers et al., 2015; Whitford & Titone, 2012, 2017a,b) and older (Whitford & Titone, 2017a,b) adults, as well as reduced L2 versus L1 text-level reading behavior in bilingual younger (Cop, Drieghe et al., 2015; Whitford & Titone, 2015, 2016) and older (Whitford & Titone, 2016) adults. Considered together, these findings suggest that the access and integration of L2 word-related information are consistently reduced in bilinguals across the lifespan. As such, even a lifetime of L2 exposure might not yield native-like reading behavior, which may have implications for the assessment of children’s reading skills in L2 learning environments (and especially in cases where the L2 represents a higher status or majority language). We note, however, that bilingual children’s L1 and L2 reading comprehension (for the experimental paragraphs) was comparable. Thus, bilingual children might also deploy additional resources and/or engage in adaptive strategies during L2 reading to offset any processing disadvantages associated with their reduced L2 experience and, ultimately, successfully complete the task at hand.

Differences between children and adults Third, we found that both global and local measures of reading performance were reduced in both groups of children versus adults (across their known languages). Global reductions included slower reading rates, longer fixation durations, more saccades (both progressive and regressive), and longer total reading times; local reductions included longer first fixation durations and gaze durations. Also with respect to local processing, both groups of children displayed larger L1 word frequency effects during both early-stage and late-stage reading; bilingual children also displayed larger L2 word frequency effects, but during early-stage reading only. The children’s larger word frequency effects (across the L1 and L2) were driven by their differentially slower processing of lower-frequency words. Interestingly, however, the differences between bilingual children’s and adults’ global and local reading performance were smaller in the L2 than in the L1 (see, e.g., Figs. 3 and 5). Although the bilingual adults necessarily have greater absolute L2 experience than the children, their current levels of L2 exposure were significantly lower (8.67% vs. 39.70%, p < .001). This suggests that between-group differences in L2 reading behavior related to greater historical (or total) experience with print (and language more generally) can be reduced by relatively low levels of current reading exposure. In contrast, the bilingual children were more actively exposed to L2 print, which also functionally reduced the magnitude of between-language differences in reading behavior. Again, these findings corroborate the weaker links hypothesis (Gollan et al., 2008, 2011) and suggest that adults’ greater absolute langauge experience reduces the effective range of word frequencies, thereby reducing differences in lexical quality and accessibility between lower-frequency and higherfrequency words. These findings are also consistent with recent (monolingual) modeling work demonstrating that age differences in the eye movement record are driven by children’s reduced rates of lexical processing (Mancheva et al., 2015; Reichle et al., 2013) and extend recent (monolingual) experimental work reporting larger word frequency effects in children than in young adults during natural reading (Blythe, Liversedge, Joseph, White, & Rayner, 2009; Tiffin-Richards & Schroeder, 2015; cf. Blythe et al., 2006).

V. Whitford, M.F. Joanisse / Journal of Experimental Child Psychology 173 (2018) 318–337

335

Lastly, although our findings provide important insights into developmental differences (and similarities) in the eye movement record, one point to consider is that the word frequency values used in the current study were obtained from adult corpora and, thus, might not accurately capture children’s exposure to words (for further discussion, see Blythe & Joseph, 2011; Joseph et al., 2013; Luke et al., 2015). Although we observed robust word frequency effects in both age groups (and across both languages) using adult estimates, our findings might not generalize to future bilingual investigations using child estimates. This concern, however, does not apply to the reading differences we observed between the monolingual and bilingual children. Conclusion The current study represents the first comprehensive investigation of online reading performance in bilingual children relative to their monolingual peers and relative to a young adult comparison group. Eye movement recordings were used to assess the reading of naturalistic texts, which yielded both word-level and text-level measures of reading performance. The findings suggest that duallanguage reading acquisition in bilingual children does indeed affect their reading performance (even when such an impact is not apparent in offline reading contexts), leading to quantitative differences in the eye movement record. Such differences should not be construed as deficits, as reading comprehension was comparable in both groups of children. Rather, they should be interpreted as processing trade-offs that are a consequence of actively encoding two language systems in the young developing brain. Acknowledgments This research was supported by a Natural Sciences and Engineering Research Council of Canada (NSERC) Operating Grant (Marc F. Joanisse), an NSERC Research Tools and Instruments Grant (Department of Psychology, University of Western Ontario), and a Fonds québecois de recherche sur la nature et les technologies (FQRNT) postdoctoral scholarship (Veronica Whitford). The funding sources had no involvement with any aspect of this article. Special thanks go to the children, parents, and administrators at the Conseil scolaire catholique Providence, Conseil scolaire Viamonde, London District Catholic School Board, and Thames Valley District School Board for making this research possible. Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at https://doi. org/10.1016/j.jecp.2018.03.014. References Aghababian, V., & Nazir, T. (2000). Developing normal reading skills: Aspects of the visual processes underlying word recognition. Journal of Experimental Child Psychology, 76, 123–150. Andrews, S., & Hersch, J. (2010). Lexical precision in skilled readers: Individual differences in masked neighbor priming. Journal of Experimental Psychology: General, 139, 299–318. August, D., & Shanahan, T. (Eds.). (2006). Developing literacy in second-language learners: Report of the national literacy panel on language-minority children and youth. Mahwah, NJ: Lawrence Erlbaum. Baayen, H. R. (2008). Analyzing linguistic data: A practical introduction to statistics using R. New York: Cambridge University Press. Baayen, H. R., Davidson, D. J., & Bates, D. M. (2008). Mixed-effects modeling with crossed random effects for subjects and items. Journal of Memory & Language, 59, 390–412. Balota, D. A., Yap, M. J., Hutchison, K. A., Cortese, M. J., Kessler, B., Loftis, B., ... Treiman, R. (2007). The English lexicon project. Behavior Research Methods, 39, 445–459. Barr, D. J., Levy, R., Scheepers, C., & Tily, H. J. (2013). Random effects structure for confirmatory hypothesis testing: Keep it maximal. Journal of Memory and Language, 68, 255–278. Bates, D., Mächler, M., Bolker, B., & Walker, S. (2014). Fitting linear mixed-effects models using lme4. arXiv preprint arXiv:1406.5823. Bates, D., Kliegl, R., Vasishth, S., & Baayen, H. R. (2015). Parsimonious mixed models. arXiv:1506.04967. Berens, M. S., Kovelman, I., & Petitto, L.-A. (2013). Should bilingual children learn reading in two languages at the same time or in sequence? Bilingual Research Journal, 36, 35–60.

336

V. Whitford, M.F. Joanisse / Journal of Experimental Child Psychology 173 (2018) 318–337

Blythe, H. I., Häikiö, T., Bertam, R., Liversedge, S. P., & Hyönä, J. (2011). Reading disappearing text: Why do children refixate words? Vision Research, 51, 84–92. Blythe, H. I., & Joseph, H. S. S. L. (2011). Children’s eye movements during reading. In S. P. Liversedge, I. Gilchrist, & S. Everling (Eds.), The Oxford handbook of eye movements (pp. 643–662). Oxford, UK: Oxford University Press. Blythe, H. I., Liversedge, S. P., Joseph, H. S. S. L., White, S. J., Findlay, J. M., & Rayner, K. (2006). The binocular co-ordination of eye movements during reading in children and adults. Vision Research, 46, 3898–3908. Blythe, H. I., Liversedge, S. P., Joseph, H. S. S. L., White, S. J., & Rayner, K. (2009). Visual information capture during fixations in reading for children and adults. Vision Research, 49, 1583–1591. Brown, L., Sherbenou, R. J., & Johnsen, S. K. (1997). Test of nonverbal intelligence, third edition: Examiner’s manual. Austin, TX: ProEd. Brysbaert, M., & New, B. (2009). Moving beyond Kucera and Francis: A critical evaluation of current word frequency norms and the introduction of a new and improved word frequency measure for American English. Behavior Research Methods, 41, 977–990. Cop, U., Drieghe, D., & Duyck, W. (2015). Eye movement patterns in natural reading: A comparison of monolingual and bilingual reading of a novel. PLoS ONE, 10(8), e0134008. Cop, U., Keuleers, E., Drieghe, D., & Duyck, W. (2015). Frequency effects in monolingual and bilingual natural reading. Psychonomic Bulletin & Review, 225, 1216–1234. Dijkstra, T., & Van Heuven, W. J. B. (2002). The architecture of the bilingual word recognition system: From identification to decision. Bilingualism: Language and Cognition, 5, 175–197. Ducrot, S., Pynte, J., Ghio, A., & Lété, B. (2013). Visual and linguistic determinants of the eyes’ initial fixation position in reading development. Acta Psychologica, 142, 287–298. Engbert, R., Nuthmann, A., Richter, E. M., & Kliegl, R. (2005). SWIFT: A dynamical model of saccade generation during reading. Psychological Review, 112, 777–813. Frey, A. (2016). Eye movements in children during reading: A review. In International symposium for educational literacy (pp. 69–90). Jouvence, Quebec, Canada: Les Éditions de l’Université de Sherbrooke. Gollan, T. H., Montoya, R. I., Cera, C., & Sandoval, T. C. (2008). More use almost always means a smaller frequency effect: Aging, bilingualism, and the weaker links hypothesis. Journal of Memory and Language, 58, 787–814. Gollan, T. H., Slattery, T. J., Goldenberg, D., Van Assche, E., Duyck, W., & Rayner, K. (2011). Frequency drives lexical access in reading but not in speaking: The frequency-lag hypothesis. Journal of Experimental Psychology: General, 140, 186–209. Green, D. A., & Riddell, C. W. (2007). Literacy and the labour market: The generation of literacy and its impact on earnings for nativeborn Canadians (International Adult Literacy Survey, No. 18). Statistics Canada Cat. No. 89–552-XIE. Häikiö, T., Bertram, R., & Hyönä, J. (2010). Development of parafoveal processing within and across words in reading: Evidence from the boundary paradigm. The Quarterly Journal of Experimental Psychology, 63, 1982–1998. Häikiö, T., Bertram, R., Hyönä, J., & Niemi, P. (2009). Development of the letter identity span in reading: Evidence from the eye movement moving window paradigm. Journal of Experimental Child Psychology, 102, 167–181. Hollingshead, A. B. (1975). Four-factor index of social status. Unpublished manuscript, Yale University. Huestegge, L., Radach, R., Corbic, D., & Huestegge, S. M. (2009). Oculomotor and linguistic determinants of reading development: A longitudinal study. Vision Research, 49, 2948–2959. Hyönä, J., & Olson, R. K. (1995). Eye fixation patterns among dyslexic and normal readers: Effects of word-length and wordfrequency. Journal of Experimental Psychology: Learning, Memory, and Cognition, 21, 1430–1440. Jared, D. (2015). Literacy and literacy development in bilinguals. In A. Pollatsek & R. Treiman (Eds.), Oxford handbook of reading (pp. 1–21). New York: Oxford University Press. Joanisse, M. F., & McClelland, J. L. (2015). Connectionist perspectives on language learning, representation, and processing. Wiley Interdisciplinary Reviews: Cognitive Science, 6, 235–247. Joseph, H. S., & Liversedge, S. P. (2013). Children’s and adults’ on-line processing of syntactically ambiguous sentences during reading. PLoS ONE, 8(1), e54141. Joseph, H. S., Liversedge, S. P., Blythe, H. I., White, S. J., Gathercole, S. E., & Rayner, K. (2008). Children’s and adults’ processing of anomaly and implausibility during reading: Evidence from eye movements. The Quarterly Journal of Experimental Psychology, 61, 708–723. Joseph, H. S. S. L., Liversedge, S. P., Blythe, H. I., White, S. J., & Rayner, K. (2009). Word length and landing position effects during reading in children and adults. Vision Research, 49, 2078–2086. Joseph, H. S., Nation, K., & Liversedge, S. P. (2013). Using eye movements to investigate word frequency effects in children’s sentence reading. School Psychology Review, 42, 207–222. Kirsch, I. S., Jungeblat, A. J., Jenkins, L., & Kolstad, A. (2002). Adult literacy in America: A first look at findings of the National Adult Literacy Survey (3rd ed.). Washington, DC: National Center for Education, U.S. Department of Education. Kuperman, V., & Van Dyke, J. A. (2013). Reassessing word frequency as a determinant of word recognition for skilled and unskilled readers. Journal of Experimental Psychology: Human Perception and Performance, 39, 802–823. La Paro, K. M., & Pianta, R. C. (2000). Predicting children’s competence in the early school years: A meta-analytic review. Review of Educational Research, 70, 443–484. Leeuw, L., Segers, E., & Verhoeven, L. (2015). Role of text and student characteristics in real-time reading processes across the primary grades. Journal of Research in Reading, 39, 389–408. Lew-Williams, C. (2017). Specific referential contexts shape efficiency in second language processing: Three eye-tracking experiments with 6- and 10-year-old children in Spanish immersion schools. Annual Review of Applied Linguistics, 37, 128–147. Liang, F., Wang, Y., Yang, W., & Bai, X. (2017). The modulation of reading level on the development of children’s eye movement characteristics: Evidence from 9- to 11-year-old children. Acta Psychologica Sinica, 49, 450–459. Luke, S. G., Henderson, J. M., & Ferreira, F. (2015). Children’s eye-movements during reading reflect the quality of lexical representations: An individual differences approach. Journal of Experimental Psychology: Learning, Memory, and Cognition, 41, 1675–1683.

V. Whitford, M.F. Joanisse / Journal of Experimental Child Psychology 173 (2018) 318–337

337

Mancheva, L., Reichle, E. D., Lemaire, B., Valdois, S., Ecalle, J., & Guérin-Dugué, A. (2015). An analysis of reading skill development using EZ Reader. Journal of Cognitive Psychology, 27, 657–676. Marian, V., Blumenfeld, K. H., & Kaushanskaya, M. (2007). The Language Proficiency and Experience Questionnaire (LEAP-Q): Assessing language profiles in bilinguals and multilinguals. Journal of Speech, Language, and Hearing, 50, 940–967. McConkie, G. W., Zola, D., Grimes, J., Kerr, P. W., Bryant, N. R., & Wolff, P. M. (1991). Children’s eye movements during reading. In J. F. Stein (Ed.), Vision and visual dyslexia (pp. 251–262). Boston: CRC Press. Miellet, S., Sparrow, L., & Sereno, S. C. (2007). Word frequency and predictability effects in reading French: An evaluation of the E-Z Reader model. Psychonomic Bulletin & Review, 14, 762–769. Monaghan, P., Chang, Y. N., Welbourne, S., & Brysbaert, M. (2017). Exploring the relations between word frequency, language exposure, and bilingualism in a computational model of reading. Journal of Memory and Language, 93, 1–21. Nation, K. (2017). Nurturing a lexical legacy: Reading experience is critical for the development of word reading skill. npj Science of Learning, 2. https://doi.org/10.1038/s41539-017-0004-7. New, B., Pallier, C., Ferrand, L., & Matos, R. (2001). Use base de donnés lexicales du français contemporain sur internet: LEXIQUE. L’Année Psychologique, 101, 447–462. Perfetti, C. A. (2007). Reading ability: Lexical quality to comprehension. Scientific Studies of Reading, 11, 357–383. Perfetti, C. A., & Hart, L. (2002). The lexical quality hypothesis. In L. Verhoeven, C. Elbro, & P. Reitsma (Eds.), Precursors of functional literacy (pp. 189–213). Amsterdam: John Benjamins. Pollatsek, A., Reichle, E. D., & Rayner, K. (2006). Tests of the E-Z Reader model: Exploring the interface between cognition and eye movement control. Cognitive Psychology, 52, 1–56. R Development Core Team. (2017). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. . Radach, R., Huestegge, L., & Reilly, R. (2008). The role of global top-down factors in local eye-movement control in reading. Psychological Research Psychologische Forschung, 72, 675–688. Rayner, K. (1986). Eye movements and the perceptual span in beginning and skilled readers. Journal of Experimental Child Psychology, 41, 211–236. Rayner, K. (1998). Eye movements in reading and information processing: 20 years of research. Psychological Bulletin, 124, 372–422. Rayner, K. (2009). Eye movements and attention in reading, scene perception, and visual search. The Quarterly Journal of Experimental Psychology, 62, 1457–1506. Rayner, K., Pollatsek, A., Ashby, J., & Clifton, C. E. (2012). The psychology of reading. New York: Psychology Press. Reichle, E. D., Liversedge, S. P., Drieghe, D., Blythe, H. I., Joseph, H. S., White, S. J., & Rayner, K. (2013). Using E-Z Reader to examine the concurrent development of eye-movement control and reading skill. Developmental Review, 33, 110–149. Reichle, E. D., Pollatsek, A., Fisher, D. L., & Rayner, K. (1998). Toward a model of eye movement control in reading. Psychological Review, 105, 125–157. Seidenberg, M. S., & McClelland, J. L. (1990). A distributed, developmental model of word recognition and naming. Psychological Review, 96, 523–568. Sperlich, A., Meixner, J., & Laubrock, J. (2016). Development of the perceptual span in reading: A longitudinal study. Journal of Experimental Child Psychology, 146, 181–201. Sperlich, A., Schad, D. J., & Laubrock, J. (2015). When preview information starts to matter: Development of the perceptual span in German beginning readers. Journal of Cognitive Psychology, 27, 511–530. Tiffin-Richards, S. P., & Schroeder, S. (2015). Children’s and adults’ parafoveal processes in German: Phonological and orthographic effects. Journal of Cognitive Psychology, 27, 531–548. Titone, D., Whitford, V., Lijewska, A., & Itzhak, I. (2016). Bilingualism, executive control, and eye movement measures of reading: A selective review and reanalysis of bilingual vs. multilingual reading data. In J. Schwieter (Ed.), Cognitive control and consequences in the multilingual mind (pp. 11–46). Amsterdam: John Benjamins. Vitu, F., McConkie, G. W., Kerr, P., & O’Regan, J. K. (2001). Fixation location effects on fixation durations during reading: An inverted optimal viewing position effect. Vision Research, 41, 3513–3533. Vorstius, C., Radach, R., & Lonigan, C. J. (2014). Eye movements in developing readers: A comparison of silent and oral sentence reading. Visual Cognition, 22, 458–485. Wechsler, D. (2005). Wechsler individual achievement test, 2nd ed. (WIAT II). London: Psychological Corporation. Whitford, V., Pivneva, I., & Titone, D. (2016). Eye movement methods to investigate bilingual reading. In R. R. Heredia, J. Altarriba, & A. B. Cies´licka (Eds.), Methods in bilingual reading comprehension research (pp. 183–211). New York: Springer. Whitford, V., & Titone, D. (2012). Second language experience modulates first- and second-language word frequency effects: Evidence from eye movement measures of natural paragraph reading. Psychonomic Bulletin & Review, 19, 73–80. Whitford, V., & Titone, D. (2014). The effects of reading comprehension and launch site on frequency–predictability interactions during paragraph reading. Quarterly Journal of Experimental Psychology, 67, 1151–1165. Whitford, V., & Titone, D. (2015). Second-language experience modulates eye movements during first- and second-language sentence reading: Evidence from a gaze-contingent moving window paradigm. Journal of Experimental Psychology: Learning, Memory, and Cognition, 41, 1118–1129. Whitford, V., & Titone, D. (2016). Eye movements and the perceptual span during first- and second-language sentence reading in bilingual older adults. Psychology and Aging, 31, 58–70. Whitford, V., & Titone, D. (2017a). Lexical entrenchment and cross-language activation: Two sides of the same coin for bilingual reading across the adult lifespan. Bilingualism: Language and Cogntition, 1–20. Whitford, V., & Titone, D. (2017b). The effects of word frequency and word predictability during first- and second-language paragraph reading in bilingual older and younger adults. Psychology and Aging, 32, 158–177.