This work was supported by a Sigma Xi dissertation grant to Elaine Reese and a Spencer. Foundation grant to Robyn Fivush. Many thanks go to Liza Dondonan,.
Cognitive
Development,
10, 381405
(1995)
Predicting Children’s literacy from Mother-Child Conversations Elaine Reese University
of Otago
Early mother-child conversations may be differentially related to children’s emergent literacy, depending on the contexts examined and the function of specific utterance types within those contexts. Twenty White middle-class mothers talked about shared past events and read books with their children at 40,46, and 58 months of age. Children completed a comprehensive sion analyses, distinct skills.
Mothers’
dicted
children’s
overall
earlier abilities.
use of demanding,
emerged for children’s decontextualized
predictor
participation
utterances positively
preutter-
predictor period, especially during talk about the past, was an of children’s
figured
70-month
more prominently
print performance.
unique contributions
Finally,
in their later narrative
ImplicaGons are discussed for scaffolding
and children’s
regres-
print and narrative
print skills. However, mothers’ increasing use of contextualizing
ances over the 18-month even stronger
literacy assessment at 70 months of age. Through
patterns of prediction
skills
models of mother-child
children’s than print
conversation
to their own narrative development.
In literate societies, children’s development of reading and writing skill is a primary accomplishment on their way to full participation in the culture. Unlike the acquisition of spoken language, at which most children achieve at least minimum competence, not all children go on to be successful readers and writers (Olson, 1984; Snow, 1983). Therefore, it is important to isolate the factors that contribute to success and failure at reading and writing. Researchers now recognize that the acquisition of literacy begins long before children enter formal schooling at age 5 or 6 (e.g., Mason & Allen,
This work was supported by a Sigma Xi dissertation grant to Elaine Reese and a Spencer Foundation grant to Robyn Fivush. Many thanks go to Liza Dondonan, Marcella Eppen, Catherine Haden, Janet Kuebli, and Laura Underwood for their assistance in data collection, coding, transcribing, and critiquing. I would also like to thank the members of my dissertation committee (Dick Neisser, Howard Rollins, Mike Tomasello, and Gene Winograd) for their careful reading and valued suggestions. I am especially grateful to Robyn Fivush for support throughout the project and to Irwin Waldman for invaluable statistical advice. Finally, I am most indebted to the mothers and children who participated in this study. Correspondence and requests for reprints should be sent to Elaine Reese, University of Otago, Psychology Department, Box 56, Dunedin, New Zealand. E-mail: . Manuscript received October 13, 1994; revision accepted February 7, 1995
381
382
Elaine Reese
1986). The focus of the study of emergent literacy has been on preschoolers’ experiences with books and other print sources (such as road signs and food boxes; for example, Heath, 1983; Teale & Sulzby, 1986) as well as on preschoolers’ interactions with parents, teachers, and siblings. Although parents’ explicit instruction may play a role in children’s reading skill (see Price, Hess, & Dickson, 1981) a more pervasive contributor to children’s literacy may be the quality of everyday conversations between parents and children. This longitudinal study examines two types of maternal conversations with preschoolers to assess contributions to children’s literacy, specifically to their reading and narrative skills. Snow (1983) has argued that adults’ decontextualization strategies in conversation may help prepare children for understanding the written word. Decontextualized talk occurs when adults focus on topics about which there are no cues in the immediate environment, thereby requiring children to go beyond perceptually present information to comprehend the spoken word. Deciphering the written word entails understanding decontextualized language to an even greater degree. Parental decontextualization can be viewed as a form of cognitive distancing, a strategy that is related to children’s representational abilities (e.g., Sigel, 1993). Just as specific utterances may differ in their degree of decontextualization, so too may conversational contexts as a whole vary in the demands placed on the child. Snow (1983) has proposed that parent-child conversations about past events are a highly decontextualized context, because in these conversations children must access events that are not in the present place or time. This study focuses on mother-child past event narratives as a predictor of children’s literacy in comparison to another everyday context for middle-class children: mother-child storybook reading. Mother-child book reading varies in degree of decontextualization as a function of the mothers’ use of probing comments about the story and the nature of print as opposed to more contextualized commentary about pictures. Arguably, mother-child past event narratives can also be highly contextualized for children, Mothers may provide a great deal of information about the target past event and leave very little memory work to the child. Questions of how much task responsibility adults place on children address the concept of scaffolding, or the degree to which adults sensitively structure tasks in accord with children’s capabilities (see Wood, Bruner, & Ross, 1976). Scaffolding can be a somewhat problematic notion when attempting to hypothesize effects on children’s literacy, primarily because theorists differ in what constitutes an appropriate adult response to children’s increased competence. One implicit view of scaffolding present in much of the emergent literacy research is that the more decontextualized the adults’ side of the conversation, the better the outcome for children’s literacy. However, many explicit conceptualizations of scaffolding, such as
Predicting
literacy
383
Vygotsky’s (1978) zone of proximal development and Rogoff’s (1990) guided participation model, depend instead on a withdrawal of adult support as the child grows more competent in the task: in Rogoff’s terms, a “transfer of responsibility.” Thus, scaffolding in this latter view consists of a decrease in contextualization as the child’s skill level increases, but doesn’t necessarily predict a simultaneous increase in decontextualization. This study addresses both of these processes. One goal is to pull apart functionally different maternal utterances, across the contextualized-to-decontextualized continuum, to examine their different relationships with child literacy. An additional goal more directly tests scaffolding notions by examining mother-child conversations across the preschool period, exploring the possibility of changing beneficial predictors over this crucial period prior to children’s entry into formal schooling. Finally, the type of scaffolding that is optimal may well be different for different child literacy skills. One intriguing possibility is the notion of different pathways to prereading aspects of literacy, such as print concepts and letter recognition, and narrative understanding, which additionally involves oral language skill (see Snow & Dickinson, 1991 for similar arguments on the different components of literacy). Specifically, acquisition of prereading skills may depend more on adult-directed than child-directed interactions (Rogoff, 1990). Previous research has examined the relationship between adult-child conversation and children’s literacy in several ways. One group of intervention studies has demonstrated effects of adult book reading on children’s language and literacy skill (e.g., Mason, Kerr, Sinha, & McCormick, 1990; Valdez-Menchaca & Whitehurst, 1992; Whitehurst et al., 1988). These interventions, however, have focused primarily on the effects of a predetermined reading style instead of examining naturally occurring variability in adult reading style as it relates to children’s literacy. Conceivably, book-reading techniques different from those chosen by the above researchers may be less or more beneficial, and in different ways, for children’s language and literacy development. In addition, these interventions were designed to be short term and did not explore the possibility of changing beneficial reading strategies across the preschool period. Another group of studies consists of longitudinal examinations of adult-child conversation in a variety of contexts as precursors to children’s literacy skill (e.g., Dickinson, 1993; Dickinson & Moreton, 1991; Dickinson & Smith, 1994). Most relevant to the present study, Beals, De Temple, Tabors, and Snow (1991) explored mother-child conversation during book reading, past event narratives, and meal time when children were aged 3 and 4 in relation to children’s emergent literacy at age 5. Beals et al. found that book reading does not appear to be the only conversational context that contributes to children’s literacy: Past event narratives and mealtime conversations were also related to a number of literacy skills. In turn, the
384
Elaine Reese
different contexts predicted quite different aspects of literacy. Finally, children’s contributions to the various conversations were a strong predictor of several later literacy skills. Although the Beals et al. (1991) study demonstrated some interesting relationships between conversational contexts and literacy skills, the coding of the conversations only broadly differentiated decontextualized utterances from other utterance types. Maternal utterances may vary along a continuum of decontextualization, not as a dichotomous distinction. In this study, mother-child conversations during book reading and past event narratives were assessed at three time points: when children were 40, 46, and 58 months old. Maternal utterances were coded for degree of decontextualization, and children’s utterances for their on-task, unique contributions. A shorter gap between the first two time points insured a more detailed look early in the preschool period, a time when children’s language development and hence their contributions to conversations may be increasing more rapidly. Only mother-child interactions were examined, because previous research has found no differences between mothers’ and fathers’ conversational styles with their preschoolers in either the book reading (Pellegrini, Brody, & Sigel, 1985) or past event narrative contexts (Reese & Fivush, 1993). At 70 months of age, children were administered a comprehensive literacy assessment of their print concepts, decoding, receptive vocabulary, story comprehension, story recall, and story production skills. At this point, none of the children had entered the first grade and only one child had completed kindergarten. Testing children’s literacy skills prior to intensive formal instruction may provide the cleanest picture of possible early home environment influences; however, we also acknowledged the possible effect of children’s preschool experiences by examining the quantity of reading instruction in daycare, preschool, and kindergarten prior to the 70-month literacy assessment. In keeping with a process view of adult-child interaction, the degree to which mother-child conversation contributed to children’s literacy was assessed through growth curve analysis (Willett, 1988). This procedure utilizes multiple data points over time to provide subjects’ average scores as well as their patterns of change over time as predictor variables, which are then entered into regression analyses to predict the dependent variables: in this case, children’s 70-month literacy scores. Predictions following from such a design rest on theories of adult scaffolding and children’s cognitive development. The first prediction was that a high overall level of maternal decontextualization would be positively associated with children’s literacy. Second, the use of growth curve analysis allowed a more direct test of maternal scaffolding hypotheses by predicting that a maternal decrease in contextualizing utterances across the preschool period would be positively associated with children’s independent literacy skills.
Predicting
385
literacy
METHOD
This study is part of a larger longitudinal study of parent-child interactions and children’s narrative and memory development. The full study consists of four time points with four sessions at each time point; however, only those tasks relevant to the present study will be discussed here.1 Children were 40 months old at the first time point, 46 months at the second, 58 months at the third, and 70 months at the final time point. Experimenters first contacted parents by telephone and told them they were interested in how much and what kinds of information children remember with different conversational partners. Thus, children’s memory abilities were the stated focus of the study. All sessions were conducted in the children’s homes by three female experimenters. Each experimenter started out with 8 families and conducted all subsequent sessions with the same families. Participants
Twenty-four White middle-class, two-parent families with 40-month-old children were recruited from county birth records in a metropolitan U.S. area for participation in this study. Two families moved out-of-state after the first time point, and two families elected not to continue participation because of time constraints, one after the first time point, and one after the third. In 11 of the remaining families, the target children were boys. Six of the girls and 8 of the boys were firstborns; 3 girls and 3 boys were laterborns. At the beginning of the study, 11 mothers worked full time outside the home, 5 worked part time (15-30 hr per week), and 4 did not work outside the home. By the 70-month time point, 9 of the mothers worked full time, 6 worked part time, and 5 were at-home mothers. All mothers had attended college, and 75% held college degrees. At the 70-month time point, 19 of the children were just about to begin kindergarten. One child had attended kindergarten in the previous year. Tasks and Coding
Mother-Child Storybook Reading. At each of the 40-, 46-, and 58month time points, experimenters provided mothers with a different unfamiliar storybook to read to their children. Mothers were informed that the focus of the book reading was children’s memory for stories and that later the children would be asked to recount the story. At the 40-month time
‘Other tasks included in the larger study were father-child and experimenter-child past event narrative sessions, a free-play session between mother and child, a story construction task in which children created stories separately with mothers and experimenters using provided toys, a self-concept assessment for children, and a photograph task in which mother and child looked at family photos.
386
Elaine Reese
point, mothers read either The Baby Blue Cat and the Whole Batch of Cookies (Fryor, 1989) or Peter3 Pock&~ (Rice, 1989); at the 46month time point, either A Lion for Lewis (Wells, 1982) or Lost! (McFhail, 1990); and at the S-month time point, either EIlen and the Goldfish (Himmelman, 1990) or Charlie Anderson (Abercrombie, 1990). Within time points, one half of the boys and one half of the girls received the first book; the remaining children were given the second book. There were no instances in which children had previous experience with the experimental book. Across time points, books were of comparable length in terms of total number of independent clauses, with approximately 50 independent clauses each. Within each time point, the two books had similar narrative structures with regard to plot development and use of dialogue. The experimenter was not present during the audiotaped book reading, which was later transcribed in fuil. All maternal and child utterances from the reading sessions that were not part of the actual text of the stories were coded into one of the following categories: a) descriptions or labels of the pictures (e.g., See? There’s Lewis going into the attic), b) predictions or inferences about the story (e.g., Why do you think he did that?), c) print talk, either about letters or words (e.g., That word is ‘indeed’) or whole book concepts (e.g., The illustrator is the one who drew the secures), and d) other, including general knowledge comments and relations to child’s own experience. Each proposition representing a new instance of a code within a conversational turn was coded. Children’s participation measure during book reading consisted of their unelicited, story-relevant comments. Thus, all child utterances that were related to the task at hand and that were not prompted by maternat questioning were induded in their participation score. Iwo raters independently coded 25% of the transcripts at each time point for reliability. At the first time point, average agreement for mother and child utterances (computed as a ratio of total agreements over total number of codes) was 91% (Cohen’s kappa = .88). At the 46 and 58-month time points, average agreement was 86% (kappa = .79) and 89% (kappa = X5), respectively. Reliability for coding children’s utterances as unelicited or elicited was 96% (kappa = .94). Each rater then coded half of the remaining transcripts. ~~~~e~C~~~~ Pass Event NurrNives. At each of the 40-, 46, and 58month time points, experimenters helped mothers select three shared, onetime past events to discuss with their children. Events were restricted to those that spanned no more than a day. In addition, experimenters discouraged discussions of birthdays and movies or plays, because these events are either routine or contain an inherent storyline. On average, the chosen events occurred about four months prior to the discussion. After seating mother and child comfortably and instructing the mother to discuss the past
Predicting Literacy
387
events as she normally would, the experimenter left the room. The past event conversations were audiotaped and transcribed in full. Mother and chiId utterances during the past event narratives were coded on four dimensions: a) elaborations, which provide new information about the event (e.g., Yeah, you swam at the beach, and didn’t you w seashells?); b) associative talk, which focuses on world knowledge (e.g., A moustache is when you only have a little hair, just opt the top of your lip.), relations to selflother (e.g., Was it like the ice cream you had at Stone mountain?), or related fantasy talk (e.g., Child: ~ Mother: , f), c) metamemory, which Oh, Z w-n you CUQ& the W~LBL focuses on the process of remembering (e.g., Z had forgotten about the clown!); and d) other. Independent clauses made up the unit of coding. Children’s participation consisted of their unique provisions and requests of memory information. Two raters independently coded 25% of the transcripts for reliability at each of the three time points. At the 40-month time point, average agreement for mother and child utterances was 88% (kappa = .85). At the 46and S-month time points, average agreement was 83.5% (kappa = .80) and 85% (kappa = .Sl). The maternal utterance types in each context are hypothesized to vary in degree of decontextualization. In book reading, descriptions and labels are directly related to the pictures or story events; predictions and inferences are relevant to the storyline but draw the child’s attention to nonobvious connections between story events; and print talk is a metacognitive comment on the characteristics of print or the process of book reading. Therefore, descriptions and labels are highly contextualized comments because environmental cues are present for understanding. Predictions and inferences are more decontextualized because they go beyond perceptually available information. Print talk is highly decontextualized because it is not about the story at all, but goes up a level to discuss the nature of storybook reading. In past event narratives, elaborations are always directly about the event under discussion and serve to contextualize the event for the child. Associative comments may also contextualize the event for the child, but like predictions and inferences during book reading, go beyond the event itself to make nonobvious connections which may be crucial for understanding. And like print talk during book reading, metamemory comments are not concerned with the specific past event at all but with metacognitive aspects of remembering. In locating these different utterance types along a contextualization continuum, the claim is that in relation to other utterance types in that same context, a particular utterance is more or less decontextualized. We cannot make claims of relative decontextualization across the book reading and past event narrative contexts because
Elaine Reese
388
a structurally similar utterance different contexts.
type may perform
different
functions in
Children’s 70-Month Literacy Assessment. All literacy measures, with the exception of the Peabody Picture Vocabulary Test-Revised (PPVT-R; Dunn & Dunn, 1981), were administered to children during one session at the 70-month time point in the following order. The PPVT-R was conducted at a separate session to guard against children’s fatigue. Concepts of Print. Clay’s (1979a) test of print knowledge, Stones, was administered at the beginning of the literacy assessment session. In this task, the experimenter reads a simple book to the child. The book contains a number of print errors. On each page, the experimenter stops to ask the child a print concept or decoding question. The test contains 24 questions in all, with questions 1 to 11 and 21 to 24 mainly tapping print concepts such as directionality of reading (left to right), the front and back of the book, and letter and word concepts. Questions 12 to 20 concentrate on letter and word identification skills. Children may get a possible total of 24 points if all questions are answered correctly. (See Clay [1979b] for a detailed description of the administration and scoring of the test). recoding. The Wide Range Achievement Test (reading portion) (Jastak & Jastak, 1978) was administered next in the standard prereading procedure. Children first name a series of letters and then attempt to read aloud simple words (such as cat or no) as the experimenter points to them on the page. The words increase in difficulty as children progress through the test. Ceiling is reached when children have missed 10 consecutive words. Children received a point for each letter correctly recognized and each word correctly pronounced. Experimenters then showed children three 8% X 11 Story Production. in. color drawings depicting a little girl and a woman engaged in various activities (from Rosy’s Garden, Ichikawa & Laird, 1990) and gave children the following instructions: Here are some pictures about a little girl and a woman. I want you to look at all the pictures very carefully and make up a whole story using all of the pictures. Your story doesn’t have to be only about what is in the pictures: you can also include stuff that is not in the pictures when you’re making up your story. So I want you to tell me what happens at the beginning of your story and all of the other things that happen.
Children were allowed as much time as needed to think of a story; then their stories were audiotaped and later transcribed for coding. The goal of the story production coding was to assess the overall structural quality of the stories. Children’s story productions received a composite score based on the dimensions of narrative structure, use of evaluative
Predicting literacy
389
devices, and episodic complexity (Labov, 1972; Peterson & McCabe, 1983). The component of narrative structure consisted of the child’s inclusion of major story elements such as an introduction and ending (1 point each); orientation statements about time, place, people, and setting (maximum of 7 points); complicating actions (maximum 2 points); climax and resolution (maximum of 5 points). Children’s use of evaluative devices in their stories (e.g., And it took a looooonng time.) garnered an extra point for each of 13 possible types (see Peterson & McCabe, 1983, for a complete list of evaluative types). Episodic complexity was comprised of the total number of propositions in the child’s story divided by the number of episodes or pictures (3). This complexity score was added to the child’s total story production score. Two raters independently scored 25% of children’s productions and achieved 94.4% agreement on all points (kappa = .93). One rater then scored the remaining transcripts. Story Comprehension. In the comprehension task, experimenters read an unfamiliar storybook to children (A Perfect Father’s Day, Bunting, 1991) and stopped at predetermined points to ask children 12 comprehension questions. The comprehension questions pinpointed three aspects of children’s story understanding: a) plot information (what happened), b) inferences (causal connections between plot events and inferring meaning from pictures), and c) facts about the world that are crucial for story understanding. Thus, these questions not only tapped children’s understanding of story events, but also their ability to connect stories to their general knowledge. Story readings were audiotaped and children’s responses transcribed for scoring. Children received one point each for correct answers on 10 of the questions, all of which requested just one piece of information. For one of the plot information questions (e.g., What are three of the things that Susie has done at the park so fur?), children received two points for providing all three events requested and one point for providing one or two events. One of the world knowledge questions requested a definition (e.g., What is a surprise?) and was deemed the most difficult of the 12 questions. If children responded to this question with a limited but partially correct answer (e.g., A present.), they received one point. Children received two points for responding with a correct definition which was stated in terms of a specific event rather than abstract terms (e.g., Let’s say Z had a birthday party. And you didn’t know. And then you’re right behind me and Z just said, “SURPRISE!“). Children received the full 3 points for this question only if they
gave a formal definition of “surprise” that incorporated the idea that one person must know something the other doesn’t (e.g., A surprise is something that people don’t know about.). The ability to give formal definitions is evidence of decontextualization skill and is highly related to children’s
390
Elaine Reese
standardized achievement scores (see Snow, Cancino, Gonzalez, & Shriberg, 1989). Immediately after reading A Perfect Father’s Day, exStory Retelling. perimenters put the book aside and requested children to “tell me everything you remember about that story, from beginning to end.” If children could not respond at all, experimenters said, “What’s the first thing that happened in the story?” After that point, experimenters simply confirmed children’s responses. When children paused in their retelling, experimenters prompted with “Anything else?” until children announced the end of their recounts. Retellings were audiotaped and transcribed for coding. Children received a composite score for their story retellings based on accuracy, narrative quality, and coherence (adapted from Mandler & Johnson, 1977; Morrow, 1989; Peterson & McCabe, 1983). First, the text of the original story was parsed into 59 propositions. Children’s retellings, which ranged from 2 to 18 propositions, were then compared to these 59 units of information. Children received one point for each mention of the gist of an information unit. Next, narrative quality of the retellings was rated by children’s use of evaluative devices (Peterson & McCabe, 1983), temporal or causal terms (e.g., in the beginning but not and or then), character introduction, and verbatim utterances. Each unique instance of each aspect of narrative quality merited one point. Thus, children’s repeated use of a single type of evaluative (e.g., “u lot of presents” and then later “a lot of cake”) would only receive one point. Finally, coherence of the retellings was scored through children’s sequencing of the events in accord with the original sequencing in the story. Thus, children’s stories in which 67% to 100% of their propositions were correctly sequenced received 2 points; if 34% to 66% of propositions were correctly sequenced, children received 1 point. Two raters independently coded 25% of children’s retellings and agreed on 92.8% of the points (kappa = .83). Reliability was 100% on parsing children’s recall into units to be compared with the original text. One rater went on to code the remaining transcripts. Assessment of Preschool Experiences. Seventeen of the 20 children attended a daycare or a preschool prior to the 70-month time point. One additional child did not attend preschool but did attend kindergarten in the year before the literacy assessment. To account for outside exposure to reading instruction, a researcher requested that the director of each school complete a questionnaire regarding the type and quantity of pre-reading instruction in their preschool curriculum. Directors indicated for a number of activities (e.g., storybook reading, explicit instruction in reading or writing, and letter recognition games) whether that activity was part of the school’s everyday curriculum and the approximate number of minutes spent on that activity per day.
391
Predicting Literacy
Results
First, children’s performance on the literacy assessment at 70 months will be presented. Next, growth curve analyses examine mothers’ and children’s contributions to children’s later literacy. There were very few instances of missing data. Tape recorder malfunctions or experimenter error eliminated 5 out of the 240 total tasks across all time points for the 20 dyads. Missing data points were conservatively dealt with by inserting the mean for that variable at that particular time point. Therefore, all analyses are based on scores for 20 mother-child dyads across all tasks. Children’s Literacy Measures
Only 4 of the 20 children were readers as measured by their ability to sound out words on the WRAT. These children correctly read about 25 words each. The 16 children who were not readers, including the one child who had attended kindergarten, displayed varying degrees of skill at individual letter recognition, but were not able to read words. Table 1 shows the means and standard deviations for readers and nonreaders on all literacy measures. The low cell size for readers does not allow statistical comparisons, but the pattern in the table suggests that readers may score higher than nonreaders on the tests of print concepts and receptive vocabulary but are similar to nonreaders on story comprehension and narrative measures. Visual examination and tests for skewness of the distributions of other literacy measures, excluding the WRAT, indicated normality. Overall, children’s mean scores on the Concepts of Print measure (M = 11.05) and the PPVT-R (M = 108.20) were within normal limits for this age group (Clay, 1979b; Dunn & Dunn, 1981). Overall means for the story comprehension (M = 10.28) and story production (M = 12.68) tasks were similar to those obtained on comparable measures with slightly younger children (see Beals et al., 1991).
Table 1. Children’s Performance
on ‘IO-Month Literacy Measures
Readers (n = 4) Literacy measure
Print concepts WRAT= PPVT-Ra Story comprehension Story retelling Story production astandard scores.
Nonreaders
(n = 16)
M
SD
M
SD
16.00 143.75 116.75 11.88 13.38 12.00
2.16 7.68 18.62 3.70 10.42 5.94
9.81 94.06 106.06 9.88 11.09 12.85
3.29 5.38 15.56 3.42 6.30 7.38
Elaine Reese
392
Intercorrelations Among Children’s Literacy Skills. Table 2 displays intercorrelations among children’s literacy skills. In accord with past research with kindergartners (Dickinson & Snow, 1987), the measures of print concepts, receptive vocabulary (PPVT-R), and story comprehension were positively correlated. In addition, there were high positive correlations among story tasks: between story comprehension and story retelling, as well as between retelling and story production. However, children’s print concepts and vocabulary skills were not related to their oral narrative skills (story retelling and production). These findings support hypotheses of distinctly different print/semantics and narrative components of emergent literacy. Predicting Children’s Literacy From Mother-Child Conversations The degree to which mothers’ and children’s talk in the past event narrative
and book reading contexts predicted children’s literacy skills at 70 months was examined through a series of regression analyses. Specifically, growth curve analysis allowed a contrast between mothers’ and children’s average use of an utterance type versus their patterns of change in use over time as predictors of children’s literacy. Growth Curve Analysis. Growth-curve analysis (Willett, 1988) utilizes measurements from multiple time points in a longitudinal design to estimate an individual’s pattern of growth on a variable. Using ordinary least squares regression, a growth trajectory is fitted to the subject’s scores on a variable over time. The midpoint of the resulting linear regression line represents the subject’s average score over the time period, and the slope provides an estimate of the individual’s pattern of change, indicating both degree and direction of change over several points in time. The first advantage of growth curve analysis is that fitting a regression line to multiple data points enables a separation of an individual’s true score from observed status (which necessarily incorporates measurement error), thus improving measurement precision over the straight-line method of connecting only two data points to measure change. The second advantage of this method is that it enables a developmental approach to studying development: an individual’s Table 2. Correlations Among Children’s Literacy Measures Literacy measure 1. Print concepts 2. WRAT 3. PPVT-R 4. Story comprehension 5. Story retelling 6. Story production
*p < .Ol, two-way.
1 1.00
2
.71* 1.00
3
.58* .25 1.00
4
.77* .37 .66* 1.00
5
6
.16 .19 .15 .42 1.00
.02 -.02 .0.5 .31 .58* 1.00
Predicting
Literacy
393
and direction of change, as opposed to a static measurement at one point in time, are used to predict behavior. Participants’ average use and change over time on the different variables were both of interest in predicting children’s literacy. Mothers’ and children’s raw scores at each of the three predictor time points were used to estimate average and change scores for each of the three maternal variables in each context (descriptions or labels, predictions or inferences, and print talk in the storybook reading; elaborations, associative, and metamemory talk in the past event narratives) and for children’s participation levels in the two contexts. Proportions of total talk were used for all measures because of varying conversation length. Although arc-sine transformations are sometimes performed on proportional data, they were not deemed necessary in this case, because the proportional variables were normally distributed and also because the categories of interest summed up to around only 65% of total talk, across time points and contexts.* To calculate average and change scores, a linear regression line was fitted to each participant’s proportion scores for each variable over time in months. The average score was the midpoint, and the change score was the linear slope of the resulting regression line. For example, mothers who showed a general increase in elaborations over time received a positive change score, mothers whose elaborations showed little change over time had a change score of around 0, and mothers who generally decreased in elaborations over time had a negative change score. Visual examination of residuals, Cook’s D, and leverage tests were performed on the average and change scores in their relationships to criterion variables, and no outliers or undue influences by particular cases were noted. In this study, growth curve analysis methods were used to calculate the predictor variables: mothers’ and children’s average use and change in use of different utterances over an l&month period. These average and change scores were then entered into multiple regression analyses to predict children’s 70-month literacy performance. Because of the bimodal score distribution of the WRAT, it was not included as a criterion measure in the regression analyses. rate
*In the book-reading task, total talk consisted of the partner’s total extratextual utterances. In the past event narratives, total talk consisted of the average length across events, because some dyads only discussed two events instead of three at some time points. Regression analyses based on frequencies may be obtained from the author. Results for the past event narrative context are markedly similar to analyses using proportions; however, frequency results for the book-reading context are somewhat different to proportional analyses due to correlations between frequency of each utterance type and total talk (i.e., mothers who talked a lot during book reading tended to use more of all utterance types). Mothers’ total talk during book reading was marginally negatively correlated to 4 out of 6 child literacy skills.
Elaine Reese
394
Although multiple regression analyses with a small sample size must be considered exploratory, precautions were taken to strengthen the validity of the analyses. First, predictor variables were based on multiple measurements and were screened stringently for deviations from normality. Second, the method of regression used is hierarchical. Hierarchical regression relies on a conceptually determined, rather than statistically derived, order of entry of the predictor variables, thus making it less likely that significant results will occur simply by chance. To facilitate interpretation of the regression analyses, zero-order correlations between average and change scores and children’s literacy are presented in Table 3. Mothe+Child Regression Models. Mother-child regression models were tested separately for average and change scores in each conversational context. The average models assess the possible cumulative influence of Table 3. Correlations Children’s Literacy
Literacv Measure
Between Mother-Child
Print Concepts
PPVT-R
Average
and Change Scores and
story Comprehension
Story Retelling
story Production
Average Model: Book Reading
Children’s participation Description/label Prediction/inference Print talk
-.32 .44* -.03
-.51** -.23 .38 .18
-.ll - .23 .46* -.14
-.04 .36 .26 .oo
-.12 .33 -.06 -.20
.05 .09 -.03 .44*
.08 .23 -.33 .46*
-.18 .08 -.ll .47*
-.20 .02 .13 .30
-.27 -.05 .30 .13
.04 .17 .33 .OO
.12 .45* .38 -.lO
.40 .14 .30 .Ol
.37 - .09 -.05 .23
.22 .03 -.29 .Ol
.35 .4@ .61** -.02
.18 .21 .29 .17
.56** .46* .37 -.Ol
.24 .04 -.17 -.41
.51** .13 -.50** -.42
-.11
Average Model: Past Event Narratives
Children’s participation Elaboration Associative Metamemory Change Model: Book
Reading Children’s participation Description/label Prediction/inference Print talk Change Model: Past Event Narratives
Children’s participation Elaboration Associative Metamemory *p < .05, two-way.
**p < .Ol, two-way.
Predicting literacy
395
mothers’ general use of utterance types over time, whereas the change models investigate the predictive value of children’s growing participation and either mothers’ adaptation to children’s changing competence or different maternal expectations as a function of children’s age. Thus, the average models test the hypothesis that the overall level of maternal decontextualization is positively related to child literacy. The change models address scaffolding hypotheses of the positive role of mothers’ decreasing contextualization over time. In each model, the relevant child variable was entered first in the regression equation (e.g., for the average book-reading model, children’s average participation during book reading was the first variable entered). Then the relevant maternal variables were entered separately, in order of increasing decontextualization. Thus, in book reading models, descriptions and labels were entered second, predictions and inferences third, and print talk fourth. In past event narrative models, elaborations were entered second, associative talk third, and metamemory comments last. The rationale for this order of entry is that it provides the most stringent test of maternal contributions, especially maternal decontextualized utterances. An additional advantage is that the children’s participation levels are controlled for, so that any relationships found for maternal variables are independent of children’s average or changing participation in that context. In presenting results from multiple regression analyses on maternal and child models, individual variables will be noted if they contributed uniquely to the regression solution once the effect of other predictor variables had been taken into account (i.e., those variables with a significant semipartial correlation upon entry). Results will also be discussed in terms of the overall pattern that the regressions present, especially the different predictive patterns obtained for print or semantics skills (print concepts, vocabulary, and story comprehension) and narrative skills (story retelling and story production). MothecChild Average Models. Table 4 contains results of the average regression models for book reading and past event narratives. During book reading, children’s average participation was significantly related only to the PPVT-R. The lower children’s average participation in preschool book reading, the higher their 70-month receptive vocabulary score. Maternal variables were slightly stronger predictors of literacy, with predictions and inferences positively predicting children’s story comprehension and story retelling. The overall pattern of mothers’ average use of different variables during book reading revealed a possible negative role for descriptions and labels and a generally positive role for predictions/inferences across all print and semantics tasks. The average book reading models as a whole were good predictors of children’s receptive vocabulary (model R* = .46) and their story retelling scores (model R* = .39).
.05 .12 -.07 .55
-.I1 -.34 .31 -.03
P
.oo .Ol .Ol .26** R2 = .28
.Ol .ll .11 SKI Rzz.23
AR2
.08 .28 -.45 .67
-.51 -.29 .31 .20
P
AR2
,2t-3***
.Ol .07 .18* 39***
.26* .08 .OS -04 R2 = .46**
PPW-R
-.18 .03 -.09 .58
-.11 -.24 .44 -.12
P
.03 .OO .Ol .29** R2 = .33
.Ol .06 .15* JlJ R-2= .23
AR2
Story Comprehension
Average Models and Children’s Literacy
Note. p weights represent coefficient upon entry into equation. *p < .lO. **p < .05. ***p < .Ol.
Children’s average participation Elaboration Associative Metamernory
Past Event Narratives
Children’s average participation Descriptionilabel Prediction/inference Print talk
Book Reading
Predictor
Print Concepts
Table 4. Predictive Relationships: Mother-Child
-.20 -.04 .20 .34
-.04 .36 .53 .18
P
.04 .OO .04 _lo R2 = .18
.OO .13 .23** _03 R2 = .39*
AR2
Story Retelling
- .27 -.15 .44 .08
-.12 .32 .lO -.13
P
.07 .02 .17* .Ol R2= .21
.Ol .lO .Ol &2 R2 = .14
AR2
Story Production
Predicting Literacy
397
In the average model for past event narratives, mothers’ average use of metamemory comments was a positive predictor of children’s print concepts, vocabulary, and story comprehension. The positive association between maternal metamemory and children’s print and semantics skills was the strongest pattern for this model. However, mothers’ average associative talk was also a significant predictor, with a high level of associative talk during past event narratives predicting children’s lower vocabulary scores and higher story production scores. The average past event narrative model, like book reading, was a good overall predictor of children’s vocabulary (model Rz = .65). In general, then, children’s average on-task participation during book reading and past event narratives was not a good predictor of their later literacy skills. And even after controlling for children’s participation in these contexts, several maternal variables still emerged as significant predictors of children’s print and semantics skills, The overall pattern across both contexts supported the hypothesis that a high average level of maternal decontextualized utterances would positively predict children’s literacy, especially the print and semantics skills. There was some suggestion that mothers’ high use of more contextualizing utterances, such as descriptions and labels and associative talk, is a positive predictor of children’s narrative skills in the story retelling and production tasks. ~o~~e~~h~~d Change Models. Table 5 contains results of regression analyses using children’s and mothers’ change scores as predictors. As in Table 4, only those predictor variables that resulted in a significant change in Rz upon entry are discussed individually. Children’s changing participation during book reading was more predictive of literacy than was their average participation in that context. Specifically, their increasing participation was a strong positive predictor of later performance on story comprehension and story retelling. Maternal change during book reading was not a strong predictor in general of children’s literacy, but did account significantly for children’s vocabulary skill. Mothers’ increasing use of ail of the variables of interest was positively related to children’s PPVT-R scores, with a model Rz of 54. Children’s increasing participation in memory conversations also emerged as a positive predictor of later performance, especially for their story comprehension and story production scores. Even so, maternal past event narrative variables still contributed uniquely to the variance in children’s literacy, most notably for children’s print concepts and story comprehension skills. For these tasks, mothers’ increasing elaborations and associative talk during past event narratives were both extremely strong predictors. Mothers’ increasing associative talk was negatively predictive, however, of children’s story production, Overall, mother-child change during past event narratives significantly predicted children’s print concepts
.04 .20 .31 ..52
P
R2
= .jo***
_02
.30***
.12 .16*
R2 = .27
.lO 2
.04
.oo
AR2
.18 .27 .24
.18
.12 .52 .35 .58
P
_03
.07
.03 .03
R2 = .16
R2
.19
.33
.56 .38
.36
.16** = .54***
.12*
.40 .26 .27
B
R2
zz
.57***
.31*** .14* .10* @
-06 R2 = .36
.16* .06 .08
AR2
Story Comprehension
.02 .25**
AR2
PPVT-R
Change Models and Children’s Literacy
Note. p weights represent coefficient upon entry into equation. *p < .lO. **p < .05. ***, < .Ol.
Past event narratives Children’s changing participation .35 .41 Elaboration Associative .56 -.14 Metamemory
Children’s changing participation Description/label Prediction/inference Print talk
Book reading
Predictor
Print Concepts
Table 5. Predictive Relationships: Mother-Child
-.17 -.37
.24 .OO
.37 .Ol -.05 .28
P
a R2 = .17
.03
.oo
.06
.14* .oo .oo 04 Rz: .18
AR2
Story Retelling
.51 .04 -.52 .03
.22 .lO -.30 -.34
B
~2
=
-00
.52***
.25***
.26** .Ol
-06 R2 = .21
.09
.05 .Ol
AR2
Story Production
q
(D
w
F
ID
3
D1
399
Predicting literacy
(model R* = .60), story comprehension (model R* = .57), and storytelling (model R* = .52). The results of the change models across contexts first indicate a positive role for children’s changing participation. As we would expect, children’s rate of increasing participation in these conversations is a fairly good predictor of their literacy development, especially in the realm of story skills. However, even when children’s rate of development is taken into account, mothers’ changing conversational strategies over the same period uniquely predict children’s literacy skills, especially in the print and semantics realm. Additional
Analyses
In another examination of this data set, Reese (1993) did not find any significant relationships between children’s 70-month literacy and maternal education, maternal work status, or children’s initial language levels. Directors at 16 of the 18 preschools children attended prior to the 70month literacy assessment participated in curriculum interviews. Five of these 16 schools had structured academic curriculums. The remaining schools reported including informal letter and word recognition activities in the curriculum, but denied explicitly teaching reading. Two of the four children classified as readers had attended an academic preschool; the other two readers attended preschools that claimed to incorporate a whole-language approach. Correlations were computed between children’s 70-month literacy measures and two preschool variables: average minutes per day spent in group storybook reading and average minutes per day in prereading activities, the latter including explicit reading instruction as well as vocabulary and letter recognition games. Extent of prereading instruction in preschool positively predicted children’s later letter/word recognition skill on the WRAT (r = .45,p = .06) but not any other literacy measure; amount of group storybook reading was not predictive of any child literacy measure.3 DISCUSSION
Results of regression analyses showed clear and fairly strong relationships between maternal conversation and children’s literacy, especially for children’s print concepts, vocabulary, and story comprehension skills. These relationships for maternal conversation obtained even after children’s participation in the conversations had been accounted for. Children’s early conversational participation, on the other hand, showed a stronger relation-
3Because of the bimodal distribution of the WRAT, Spearman’s rho was also calculated for all correlations listed in Table 2, and for correlations between preschool variables and child literacy. These correlations showed exactly the same pattern as was obtained with Pearson’s r, with the exception that the correlation between preschool reading instruction and the WRAT reached significance (rho = .55,p < .05).
400
Elaine Reese
ship to their later story understanding and narrative abilities than to their print and vocabulary skills. These relationships between mother-child conversation and a broad range of children’s literacy skills do not appear to be attributable simply to academic instruction in preschool. Reading instruction in preschool positively predicted children’s letter/word recognition on the WRAT, but did not predict any other child literacy skill (c.f. Dickinson & Smith, 1994 for a study of the relationship between qualitative aspects of preschool instruction and children’s literacy). Along with the patterns found for maternal conversational predictors, this result is a critical indicator that exposure to explicit reading instruction may be important for decoding, but that a much broader range of semantic and narrative aspects of literacy may be acquired through incidental learning in the home prior to school entry. The obtained patterns of maternal and child contributors to literacy provide good support for hypotheses of the separability of early literacy into print, semantic, and narrative skills (e.g., Snow & Dickinson, 1991). First, children’s literacy skills correlated into two distinct clusters: 1) print and semantics (print concepts and receptive vocabulary), and 2) narrative (story retelling and story production). There were no significant or moderate correlations across these two skill groups, although the story comprehension measure was moderately correlated with both dimensions. Story comprehension may indeed tap both of these components of literacy. Second, and most important, patterns of maternal and child conversational contributors were fairly consistent within, but distinctly different across these two dimensions. For instance, a number of maternal utterance types in the two contexts were positively related to children’s print concepts, vocabulary, and comprehension, but often the same utterance type was either unrelated or negatively related to children’s narrative skills in the retelling and production tasks. Traditionally, studies of home contributors to children’s reading skill have tended to focus only on decoding aspects of literacy (see Scott-Jones, 1986 for a review). The present findings underscore the plurality of cognitive skills that make up literacy. Present findings also highlight the importance of context in the search for conversational contributors to children’s literacy. Although the book reading context was a moderate predictor of children’s vocabulary and story retelling skills, it was not the overwhelmingly strong predictor of children’s literacy many researchers and educators assume it to be. Scarborough and Dobrich (1994), in their comprehensive review of research on the role of shared book reading in children’s literacy, come to a similar conclusion about the predictive power of book reading with preschoolers. In contrast, mother+hild conversation during shared past event narratives was a more powerful predictor than book reading, especially for children’s print and semantics skills. These conversations, which have nothing
Predicting Literacy
401
to do with storybooks or print, nevertheless were the strongest predictor of children’s print concepts, vocabulary, and story comprehension. The decontextualized nature of these interactions may partially account for this relationship to those literacy skills which require a “meta” view of books and stories, but when the specific utterance types that were related are examined more closely, it appears that a different force may be at work. Although mothers’ highly decontextualized metamemory comments were a good overall predictor of children’s print and semantics tasks, even stronger relationships were evident for mothers’ increasing use of contextualizing and moderately contextualizing utterances (elaborations and associations) as these preschoolers approached formal schooling. Elaborations and associations during shared past event narratives provide additional information about the memory or link the memory with other personally significant events for the child. It may be that these contextualizing practices in a personally meaningful context especially facilitate children’s acquisition of the knowledge that words, whether spoken or written, contain meaning. ‘Ihis mapping of meaning to the written word is a crucial step in children’s emerging literacy. Bruner (1990) eloquently argues for the importance of personally significant narratives in the child’s acquisition of meaning. This study cannot address causal relationships between early reminiscing and child literacy. An alternative hypothesis, which remains to be tested, is that maternal past event narrative style is simply indicative of a more general style of maternal engagement and encouragement of children’s participation that accounts for the obtained relationship with children’s print and semantics abilities. Bus and van IJzendoorn’s (1988) finding that secure mother-child attachment styles at 24 months positively predicted children’s interest in literacy 3 years later seems relevant in this regard. These findings can, however, illuminate current conceptualizations of adult-child scaffolding and children’s cognitive development. Contrary to many models of adult-child scaffolding (see Rogoff, 1990) the obtained pattern of maternal contributors in the change models did not support the notion that a decrease in contextualizing utterances was a positive predictor of children’s later skill levels, at least for print and semantics tasks. Although a certain basic level of maternal decontextualized utterances positively predicted children’s print and semantics skills (as seen in the relationship for predictions and inferences and metamemory in the average models), it is an increasing, not decreasing, level of contextualized and moderately contextualized comments over the preschool period that positively predicts these skills. The classic view of scaffolding, however, did appear to be more in evidence for children’s narrative performance. On the story retelling and production tasks, children’s increasing participation in earlier conversations was a positive predictor whereas mothers’
402
Elaine Reese
increasing contextualization (associative talk) was negatively predictive of storytelling. In combination, these findings suggest that at the beginning of the preschool period, print and semantics skills may be promoted when the child’s role is primarily that of listening and taking direction from the adult. For these tasks, beneficial maternal scaffolding appears to consist of increased collaboration and dialogue over time as opposed to an increasing maternal emphasis on independent skills. A tentative conclusion that can be drawn from this study is that children who participate to a greater degree in early adult-child conversations may be honing different abilities from those children who allow the adult to direct the interaction. Growth curve analysis, then, may be a promising tool in the evaluation of scaffolding hypotheses, and more generally in other areas of devetopmental research. Future studies with more numerous data points could utilize nonlinear patterns of growth as predictors. Nonlinear growth patterns may indeed be the rule and not the exception for many areas of cognitive development (Bates & Carnevale, 1993). In addition, with a larger sample size, the predictive value of average-change interactions could be evaluated. That is, the same amount of maternal change may be differentially predictive of children’s skills depending on mothers’ average use of a variable. So, for exampfe, mothers who are generally high on a variable and who also increase in that variable over time may have a greater impact on certain skills than mothers who are generally low on that variable but increase their use at the same rate as the first group of mothers. The present findings, however, point almost as much to the role of the child herself as to the part of the mother in children’s literacy development. The importance of children’s early engagement for their later literacy found in the current study is in line with previous research (Beals et al., 1991; Crain-Thoreson & Dale, 1992; Wells, 198.5) and brings up the question of children’s unique influences on their own literacy development, a question of increasing interest to many researchers (see Pellegrini & Galda, 1993; Scarborough & Dobrich, 1994; Wolf & Heath, 1992). What early individual differences drive children along different pathways to literacy? In a retrospective study of 4-year-old readers’ and nonreaders’ early activity preferences, Thomas (1984) showed that readers had early and consistent preferences for quiet, prereading activities such as letter and word games, whereas nonreaders had an early and increasing preference for active, fantasy play, The current study provides suggestive evidence that early readers are not necessarily better narrators than nonreaders. Thomas’ findings reinforce the notion that decoding ability is a very specific aspect of reading acquisition, but also in conjunction with the present results, suggest the intriguing idea that children’s preference for fantasy play may propel them toward narrative, not necessarily toward “reading.” Exploring this narrative
Predicting
403
literacy
pathway to literacy is vital because narrative skill is an excellent predictor of later language outcome (Bishop & Edmundson, 1987) and school achievement (Feagans & Appelbaum, 1986). This study represents a first step in applying analyses of change to discovering the role of mother-child conversations in children’s emergent literacy. Findings underscore the importance of collaborative scaffolding between mother and child for some aspects of child literacy, such as print knowledge, vocabulary, and story comprehension, but also highlight the important role that children play in their own narrative development. Thus, the search for possible adult influences on children’s literacy should be tempered by the continued exploration of children’s unique contributions to their own growth in becoming readers and writers. REFERENCES Abercrombie, B. (1990). Charlie Anderson. New York: Margaret K. McElderry Books. Bates, E., & Carnevale, G.F. (1993). New directions in research on language development. Developmental
Review, 13,436470.
Beals, D.E., De Temple, J.M., Tabors, F.O., & Snow, C.E. (1991, April). Reading, reporting and repast: Three Rs for co-constructing language and literacy skills. Paper presented at the Biennial Meeting of the Society for Research in Child Development, Seattle, WA. Bishop, D.V.M., & Edmundson, A. (1987). Language-impaired 4-year-olds: Distinguishing transient from persistent impairment. Journal of Speech and Hearing Disorders, 52,156-173. Bruner, J. (1990). Acts of meaning. Cambridge: Harvard University Press. Bunting, E. (1991). A perfect Father’s Day. New York: Clarion Books. Bus, A., & van IJzendoorn, M. (1988). Attachment and early reading: A longitudinal study. Journal of Genetic Psychology, 149,199-210.
Clay, M.M. (1979a). Stones. Exeter, NH: Heinemann Educational Books Clay, M.M. (1979b). The early detection of reading difficulties:A diagnostic survey with recovery procedures. Exeter, NH: Heinemann Educational Books. Crain-Thoreson, C., & Dale, P.S. (1992). Do early talkers become early readers? Linguistic precocity, preschool language, and emergent literacy. Developmental Psychology, 28, 421-429. Dickinson, D.K. (1993, March). Extended language opportunities in preschool as supports for development of emergent literacy. Paper presented at the Biennial Meeting of the Society for Research in Child Development, New Orleans, LA. Dickinson, D.K., & Moreton, J. (1991, April). Predicting specific kindergarten literacy skillsfrom three-year-olds’ preschool experiences. Paper presented at the Biennial Meeting of the Society for Research in Child Development, Seattle, WA. Dickinson, D.K., & Smith, M.W. (1994). Long-term effects of preschool teachers’ book readings on low-income children’s vocabulary and story comprehension. Reading Research Quarterly, 29,105-122.
Dickinson, D.K., & Snow, C.E. (1987). Interrelationships among prereading and oral language skills in kindergartners from two social classes. Early Childhood Research Quarterly, 1, l-25.
Dunn, L.M., & Dunn, L.M. (1981). Peabody Picture Vocabulary Test-Revised. Circle Pines, MN: American Guidance Service. Feagans, L., & Appelbaum, MI. (1986). Validation of language subtypes in learning disabled children. Journal of Educational Psychology, 78,358-364.
404
Elaine Reese
Heath, S.B. (1983). Ways with words: Language, life and work in communities and classrooms. New York: Cambridge University Press Himmelman, J. (1990). Ellen and the goldfish. New York: Harper & Row. Ichikawa, S., Bc Laird, E. (1990). Rosy’s garden. New York: Phifomd Books Jastak, J., & Jastak, S. (1978). Wide-Range Achievement Test. Wilmington, DE: Jastak & Associates, Inc. Labov, W. (1972). Language in the inner city. Philadelphia: University of Pennsylvania Press Mandler, J., & Johnson, N. (1977). Remembrance of things parsed: Story structure and recall. Cognitive PsychoZogy, 9,111-l%. Mason, J.M., & Allen, J. (1986). A review of emergent literacy with implications for research and practice in reading. Review of Researchin Education, I3,3-47. Mason, J.M., Kerr, B.M., Sinha, S., & McCormick, C. (1990). Shared book reading in an Early Start program for at-risk children. Yearbook of the National Reading Conference, 39, 189-198. McPhail, D. (1990). Lost!. New York: Little, Brown & Co. Morrow, L.M. (1989). Literacy developmenf in the early years. Englewood Cliffs, NJ: Prentice Hall. Olson, D. (1984). “See! Jumping!“: Some oral language antecedents of literacy. In H. Goelman, A.A. Oberg, & E Smith (Eds), Awakening to Ziieracy (pp. 185-192). Exeter NH Heinemann Educational Books Pellegrini, A.D., Brody, G.H., & Sigel, I.E. (1985). Parents’ book-reading habits with their children. Journal of Educational Psychology, 77,332-340. Pellegrini, A.D., & Galda, L. (1993). Ten years after: A reexamination of symbolic play and literacy research. Read~g Research ~uarfer~~ 28,163-175. Peterson, C., & McCabe, A. (1983). Developmental psycholinguist~s: Three ways of Zooking at a child’s narrative. New York: Plenum. Price, G.G., Hess, R.D., & Dickson, W.P. (1981). Processes by which verbal-educational abilities are affected when mothers encourage preschool children to verbalize. Developmental Psychology, I7,554-564. Pryor, A. (1989). The baby blue cat and the whole batch of cookies. NY Viking Kestrel. Reese, E. (1993). Predicting emergent Ziteracyfrom mother~hi~d conversations ~nteructions. Unpublished doctoral dissertation, Emory University, Atlanta, GA. Reese, E., & Fivush, R. (1993). Parental styles of talking about the past. Developmental Psychology, 29.596606. Rice, E. (1989). Peter’s pockets. New York: Greenwillow Books. Rogoff, B. (1990). Apprenticeship in thinking: Cognitive development in social context. New York: Oxford University Press. Scarborough, H.S., & Dobrich, W. (1994). On the efficacy of reading to preschoolers. Developmental Review, 14,245-302. Scott-Jones, D. 91986). Family influences on cognitive development and school achievement. Review of Research in Education, II, 259-304. Sigel, I.E. (1993). The centrality of a distancing model for the development of representational competence. In R.R. Cocking & K.A. Renninger (Eds), The development and meaning ofpsychological distance (pp. 141-158). Hillsdale, NJ: Erlbaum. Snow, CE. (1983). Literacy and language: Relationships during the preschool years. Harvard Educational Review, 53, X65-189. Snow, C.E., Cancino, H., Gonzalez, P, % Shriberg, E. (1989). Giving formal definitions: An oral language correlate of literacy. In D. Bloome (Ed.), Classrooms and literacy. Norwood, NJ: Ablex. Snow, C.E., & Dickinson, D.K. (1991). Skills that aren’t basic in a new conception of literacy. In
Predicting
405
Literacy
E.M. Jennings & A.C. Purves (Eds), Literate systems and individual lives: Perspectives on literacy and schooling (pp. 179-191). Teale, W.H., & S&by, E. (Eds) (1986). Emergent literacy: Writing and reading. Norwood, NJ: Ablex. Thomas, B. (1984). Early toy preferences of four-year-old readers and nonreaders Child Development, 55,424-430.
Valdez-Menchaca, M.C., & Whitehurst, G.J. (1992). Accelerating language development through picture book reading: A systematic extension to Mexican day care. Developmenrul Psychology, 28(6), 1106-1114. Vygotsky, L.S. (1978). Mind in society: The development of higher psychological processes. In M. Cole, V. John-Steiner, S. Scribner, & E. Souberman (Eds). Cambridge: Harvard University Press Wells, G. (1985). Preschool literacy-related activities and success in school. In D.R. Olson, N. Torrance, & A. Hildyard (Eds), Literacy, language and learning: The nature and consequences of reading and writing (pp. 229-255). New York: Cambridge University Press. Wells, R. (1982). A lion for Lewis. New York: The Dial Press. Whitehurst, G.J., Falco, FL., Lonigan, C.J., Fischel, J.E., DeBaryshe, B.D., Valdez-Menchaca, MC., & Caulfield, M. (1988). Accelerating language development through picture book reading. Developmental Psychology, 24,552-559. Willett, J.B. (1988). Questions and answers in the measurement of change. Review of Research in Education,
15,345-422.
Wolf, S.A., & Heath, S.B. (1992). The braid of literature: Children’s worlds of reading. Cambridge, MA: Harvard University Press Wood, D., Bruner, J.S., & Ross, G. (1976). The role of tutoring in problem solving. Journal of Child Psychology
and Child Psychiatry,
17,89-100.