Can Babies Learn to Read? A Randomized Trial of Baby Media

Susan B. NeumanNew York University

Tanya KaeferLakehead University

Ashley PinkhamUniversity of Michigan

Gabrielle StrouseUniversity of Toronto

Targeted to children as young as 3 months old, there is a growing number of baby media products thatclaim to teach babies to read. This randomized controlled trial was designed to examine this claim byinvestigating the effects of a best-selling baby media product on reading development. One hundred andseventeen infants, ages 9 to 18 months, were randomly assigned to treatment and control groups. Childrenin the treatment condition received the baby media product, which included DVDs, word and pictureflashcards, and word books to be used daily over a 7-month period; children in the control condition,business as usual. Examining a 4-phase developmental model of reading, we examined both precursorskills (such as letter name, letter sound knowledge, print awareness, and decoding) and conventionalreading (vocabulary and comprehension) using a series of eye-tracking tasks and standardized measures.Results indicated that babies did not learn to read using baby media, despite some parents displayinggreat confidence in the programs effectiveness.

Keywords: early literacy, reading, baby media, eye tracking, vocabulary

There has been an explosion of new media targeted to babies inrecent years (Rideout & Hammel, 2006). Ignited by the 1995release of Brainy Baby, followed by a deluge of other videos andDVDs such as Baby Einstein, educational media specially mar-keted to families of infants and toddlers has become big business,with the Baby Einstein brand alone selling over $200 millionworth of products (Robb, Richert, & Wartella, 2009). A substantialproportion of these new media claim to promote infants cognitivedevelopment and vocabulary and feature testimonials and adver-tisements about how young children may benefit from these com-mercial products. Garrison and Christakis (2005), for example,reported that of the top 100 best-selling baby DVDs on Amazon.com, 76 claim to produce specific developmental benefits. Ac-cording to recent reports, these claims appear to be reaching anincreasingly receptive audience: The average 6-month-old is saidto own at least four DVDs (Barr, Danziger, Hilliard, Andolina, &

Ruskis, 2009), with this figure almost doubling by the time infantsare 18 months old (Linebarger & Vaala, 2010).

A small proportion of the makers of these baby media, however,go beyond promoting developmental benefits to argue for theirproducts ability to teach babies to read. Claiming that all babiesare Einsteins, the manufacturers of Intellectual Baby, for exam-ple, make the case that babies can learn to read beginning at age 3months (Intellectual Baby, 2009). Similarly, recognizing that ba-bies are linguistic geniuses, the makers of Brill Baby (kids arebrilliant; BrillKids, 2011) promote using their Little Readers andLittle Musicians series starting at infancy. Other products, likeYour Baby Can Read (www.ybcr.com; Titzer, 2010), claim thattoddlers will be able to read Charlottes Web and Harry Potter ifregularly exposed to their program as early as age 3 months.Teaching your baby to read is easy, one product claims (Doman& Doman, 2010). It depends on the brains ability to integrate itsvisual, auditory, linguistic and conceptual centers. And that speeddepends a great deal on the myelination of the neurons axons. . . the more myelin sheathes the axon, the faster the neuron canconduct its charge. In short, the earlier we can teach babies to read,the better (Intellectual Baby, 2009).

Although skeptics abound (American Academy of Pediatrics,1999), a small number of empirical studies have begun to examinethis assumption, targeting their focus specifically on word learn-ing. Vandewater (2011), for example, in a randomized experiment,reported that infants exposed to Baby Wordsworth for 1 monthshowed greater gains in receptive vocabulary at a 3-monthfollow-up compared with controls, although their assessment ofgains relied on parent report rather than direct assessments withchildren. In contrast, Robb et al. (2009) found no greater gains inreceptive and expressive vocabulary for 12- and 15-month-oldswho viewed the same product for 6 weeks versus a nonviewing

Susan B. Neuman, Teaching and Learning Department, SteinhardtSchool of Culture, Education, and Human Development, New York Uni-versity; Tanya Kaefer, Department of Education, Lakehead University;Ashley Pinkham, School of Education, University of Michigan; GabrielleStrouse, Department of Applied Psychology and Human Development,University of Toronto.

This research was conducted at the University of Michigan. We grate-fully acknowledge the Ready to Learn research team and the children andteachers who participated in the study.

Correspondence concerning this article should be addressed to Susan B.Neuman, Teaching and Learning Department, Steinhardt School of Cul-ture, Education, and Human Development, New York University, 239Greene Street, New York, NY 10003. E-mail: sbneuman@nyu.edu

This

docu

men

tis

copy

right

edby

the

Am

eric

anPs

ycho

logi

calA

ssoc

iatio

nor

one

ofi

tsal

lied

publ

isher

s.Th

isar

ticle

isin

tend

edso

lely

fort

hepe

rson

aluse

oft

hein

divi

dual

use

ran

dis

not

tobe

diss

emin

ated

broa

dly.

Journal of Educational Psychology 2014 American Psychological Association2014, Vol. 106, No. 2, 000 0022-0663/14/$12.00 DOI: 10.1037/a0035937

1

control group. Similarly, DeLoache and her research team (DeLo-ache et al., 2010) reported no difference in vocabulary gains for12- to 18-month-old children who viewed the same product severaltimes a week over a months time versus a control group.

Nevertheless, there are limitations to these studies that might becontested by program developers. For one, many of these productsrecommend a longer duration of treatment than multiple exposuresover 1 months time. Your Baby Can Read, for example, suggeststhat parents are likely to experience the benefits for their child onlyafter 37 months of daily use (Titzer, 2010). Second, many top-selling products include more than DVDs; for example, Intellec-tual Baby includes flashcards and books that accompany the pro-grams DVD. And third, to date, studies have only measured gains inreceptive and expressive vocabulary development. Reading profi-ciency, as we later define, involves more than oral word learning.

Therefore, one could make a case that claims about babiesabilities to learn to read have not been subject to rigorous empiricaltesting. However, some might argue that the question is moot,given the reported difficulty that infants and toddlers have learningfrom screen media (Anderson & Pempek, 2005). For example, onereason young children may struggle to learn from video presenta-tions is that they do not display dual representation, or the under-standing that pictures and videos depict not only objects in and ofthemselves but also represent similar objects in their world (De-Loache, 2004). Development of dual representation helps supportchildrens generalization and transfer of content from screen pre-sentations to other situations. Linebarger and Vaala (2010) arguedthat learning and extending new vocabulary, in particular, may beespecially dependent on dual representation, as this skill requiresinfants to simultaneously represent an object on screen both as itsown entity and as a symbol for similar objects in other environ-ments.

A substantial amount of research suggests that infants andtoddlers may not learn information as readily from screen media asfrom a live situation, a phenomenon known as the video deficit(Anderson & Pempek, 2005). The deficit has been found using avariety of language outcomes, including phonetic learning (Kuhl,Tsao, & Liu, 2003), connecting actions with a novel word (Rose-berry, Hirsh-Pasek, Parish-Morris, & Golinkoff, 2009), and iden-tifying object labels (Krcmar, Grela, & Lin, 2007; Troseth,Strouse, Verdine, & Saylor, 2013). Nevertheless, despite youngchildrens inefficient learning from video, studies that have com-pared learning from video instruction with no instruction havesuggested that infants and toddlers are capable of learning infor-mation from video (e.g., Barr & Hayne, 1999; Strouse & Troseth,2008), especially with supportive scaffolds such as co-viewing andrepeated viewings (Barr, Muentener, Garcia, Fujimoto, & Chvez,2007; Strouse & Troseth, 2013). In addition, Rice (1983) identifieda number of supportive features of videos themselves including theuse of predictable program formats, recasts, simple sentences, andslow rates of speech, some of which are prevalent in baby mediaproducts.

Despite varied research findings, many parents retain the beliefthat their children benefit from watching television. For example,a recent survey indicated that 40% of mothers of young childrenbelieve that their infants and toddlers are learning from screen time(Rideout, 2007). Therefore, it is conceivable that certain precursorsof reading (e.g., phonological awareness) are developing but maynot be apparent due to limitations in the methods traditionally used

to assess early learning. Consequently, this article was designed totake marketers at their word and to measure the effects of babymedia using a more comprehensive model of reading. Conductinga year-long randomized controlled trial in which we comparedfamilies who used a baby media product with relatively highfidelity with control families, we examined the effects of babymedia on babies reading development.

What Is Reading?Critical to a study of reading development is the very definition

of what is reading? We use the widely accepted definition ofreading reported in government consensus documents as well as ina convergence of studies (National Early Literacy Panel, 2008;National Reading Panel, 2000). Reading is a complex cognitiveprocess of decoding symbols for the intention of deriving meaningfrom print. Although it has its detractors, the simple view mostsuccinctly characterizes the process (Gough & Tunmer, 1986):Reading with understanding is the product of decoding and com-prehension, or the simple formula, R ! D " C. In actuality,however, this definition is hardly simple because it suggests amultiplicative effect: Decoding by itself is not reading; similarly,comprehension of words without the ability to unlock words intotheir constituent parts is not reading. Both must work in concert forindividuals to be able to read with meaning.

This definition contrasts with those who might argue that iden-tifying words in context such as McDonalds or Stop in stopsigns (otherwise known as environmental print) are indicators ofreal reading (Goodman, 1984). However, two independent stud-iesMasonheimer, Drum, and Ehri (1984) and Stahl and Murray(1993)have shown that while young children as early as age 2years could readily and accurately identify many logos, they couldnot read the embedded words when they were removed from thelogos. Although these behaviors, sometimes described as pseudo-reading or pretend reading (Teale & Sulzby, 1989), may highlightchildrens interest or awareness of symbol systems in their envi-ronment, they do not constitute reading or the ability to read wordsaccurately in isolation or in text with meaning.

Nevertheless, the development of reading ability is not an all-or-nothing phenomenon. Rather, researchers agree that there aredevelopmental phases that emerge or change based on internalcauses, such as developing cognitive or linguistic capabilities, andon external environmental conditions, such as the scaffoldingkinds of adult activities that support its development. In fact, thereis substantial agreement among theories of reading development(Chall, 1983; Ehri, 1979; Mason, 1980) that children move fromcontextual dependency to early decoding, where they becomeglued to print by processing letters and sounds in an effortfulmanner, ultimately to fluent and conventional reading. Such de-velopmental theories can provide researchers with a basis forassessing development and for predicting what can be expected inthe developmental path toward reading.

Consequently, our analysis of whether babies can learn to readwas based on the premise that there are developmental skills thatact as precursors to reading proficiency. Using this logic, wewould presume that even if babies could not read with understand-ing as a result of the baby media program (i.e., as indicated byprogram developers claims) they might at least exhibit some oftheir earlier skills that could accelerate later reading development.

This

docu

men

tis

copy

right

edby

the

Am

eric

anPs

ycho

logi

calA

ssoc

iatio

nor

one

ofi

tsal

lied

publ

isher

s.Th

isar

ticle

isin

tend

edso

lely

fort

hepe

rson

aluse

oft

hein

divi

dual

use

ran

dis

not

tobe

diss

emin

ated

broa

dly.

2 NEUMAN, KAEFER, PINKHAM, AND STROUSE

Consistent with this thesis, therefore, in this article, we examinereading developmentally from the emerging to the consolidationphases of reading.

Method

Participants

Participants were 117 infants, ages 1018 months (M ! 14.25months), and their parents from a small Midwestern city and thesurrounding county. These families were recruited through a va-riety of sources, including flyers distributed in the community(e.g., churches, day care centers), displays at community events(e.g., public libraries, farmers markets), social networking (e.g.,Facebook), and word of mouth. We used the following inclusioncriteria for participation in the study: (a) Babies were born full-term (i.e., # 37 weeks gestation), (b) heard English as theirprimary language (i.e., $ 10% exposure to languages other thanEnglish), and (c) did not have a history of vision, hearing, orcognitive disabilities. Together, our sample included 61 male and56 female infants, representing 88% White, 3% African American,and 9% bi/multiracial groups.

Infants were randomly assigned to one of two conditions (treat-ment or control) within counterbalanced gender and age (threeclusters: 10.012.9; 13.015.9; and 16.018.9 months) brackets.The distribution of infants in each condition did not vary acrossdemographic characteristics (all ps # .05). See Table 1 for demo-graphic information.

The majority of infants were from middle-class, highly educatedparents. Half of the families earned an annual income larger than$76,000 and 16% earned less than $30,000. Seventy-eight percentof mothers and 75% of fathers had at least a bachelors degree.Ninety-two percent of parents were married (single: 6.6%; domes-tic partners: 1.1%).

Almost half of the infants were first born and attended someform of day care at the start of the study: 13% in their own home(e.g., looked after by a grandparent), and 34% went to a day carecenter). The quality of infants home environments was relativelyhigh: 11% of families got the maximum score (i.e., 45 points) onthe Infant/Toddler Home Observation for Measurement of theEnvironment Inventory (Caldwell & Bradley, 2003). Another 83%of the families scored close to the maximum score (i.e., 3944points). Bayley scores in cognition, language, and socialemotional development were in the average range (Bayley, 2006).

Intervention MaterialsThe intervention included a best-selling baby media product,

Your Baby Can Read, sold at popular chain stores (e.g., Walmart,Target) as well as through online vendors. Marketed to childrenages 3 months and older, it purports to teach babies how to readwith fluency and comprehension within 36 months of regularuse. The intervention is composed of five volumes or units; eachvolume includes a specific number of words to be learned, rangingfrom 20 to 27 words. In Volume 1, for example, babies areexpected to learn 22 written words, ranging from two to eight

Table 1Demographic Characteristics of Treatment and Control Children (N ! 117)

Characteristic Treatment (n ! 61) Control (n ! 56) pa

Age (in no. of months)Initial visit 14.28 (2.60) 14.21 (2.70) .887Final visit 22.04 (2.81) 21.46 (2.91) .294

Gender (%) .236Male 57.4 46.4Female 42.6 53.6

Ethnicity (%) .194White 83.6 92.9African-American 4.9 Bi-/multiracial 11.5 7.1

Language exposure (%) .676English only 81.7 78.6Multiple 18.3 21.4

Siblings (%) .849Only child 47.5 43.61 39.0 40.02% 13.6 16.3

Attends child care 58.3 44.6 .140Parental education (Mdn) Bachelors degree Bachelors degree .283Household income (Mdn) $76,000$100,000 $51,000$75,000 .825Infant/Toddler HOME score 41.69 (2.74) 41.69 (3.29) .215Bayley Scales percentile

Cognitive 56.36 (23.19) 51.28 (24.43) .251Language 42.07 (25.70) 40.38 (23.80) .714Socialemotional 48.99 (28.02) 57.70 (28.43) .100

Note. Standard deviations are presented in parentheses; HOME ! Infants and Toddlers Version of the HomeObservation for Measurement of the Environment inventory (Caldwell & Bradley, 2003), maximum score! 45;Bayley Scales ! Bayley Scales of Infant and Toddler Development (3rd ed.; Bayley, 2006).a p reported for t test or chi-square test of treatment versus control groups.

This

docu

men

tis

copy

right

edby

the

Am

eric

anPs

ycho

logi

calA

ssoc

iatio

nor

one

ofi

tsal

lied

publ

isher

s.Th

isar

ticle

isin

tend

edso

lely

fort

hepe

rson

aluse

oft

hein

divi

dual

use

ran

dis

not

tobe

diss

emin

ated

broa

dly.

3RANDOMIZED TRIAL OF BABY MEDIA

letters in length. According to the documentation included with theintervention, words are selected carefullyalthough no rubric forselection is provided. Across the different materials, similar wordsare presented in different fonts and different colors.

Each volume of the intervention includes a DVD, word cards,picture cards, and a picture book. According to the parent guideincluded with the intervention, infants should watch the DVD atleast one time per day and devote at least 15 min with each of theother three materials.

DVDs. Each 20-min DVD volume begins with a brief intro-ductory segment in which infants are specifically instructed toattend to the text appearing on-screen. DVDs then follow a rou-tinized format. Lowercase printed words (e.g., ears) are showncentered against a solid colored background, taking approximatelyone third of the visual space. After a 2-s delay, an adult voiceoversays a familiar carrier phrase (e.g., Can you say [target word])followed by a childs voice saying the word that appears on screen.After approximately 2 s, a second child voiceover repeats theonscreen word. Each time the printed word is spoken, a cursorappears beneath the word and travels the length of the word fromleft to right. The printed word then disappears and is replaced witha scene depicting the meaning of the word (e.g., a child pointing toher ears) accompanied by an adult voiceover describing the scene(e.g., Katie is pointing to her ears). For each volume, this pattern(i.e., printed word followed by scene depicting the word) occursbetween 60 and 70 times, with each of the target words presentedonly one time or as many as five times. Each DVD also includesthree brief songs or nursery rhymes (e.g., The Itsy Bitsy Spider).

Word cards. Each volume includes a set of 1012 flashcardsmeasuring 7.5 by 4.5 in. One word is printed on each side of thecard (i.e., 2024 words total). The words are printed in all low-ercase letters in black ink on a colored background; each word iswritten in a distinct font.

Picture cards. Each volume includes five picture cards. Pic-ture cards measure 7.5 by 4.5 in. and have one word written oneach side (i.e., 10 words total). Words are printed in lowercaseletters in black ink; font and background color vary from word toword. A 1.5-in. tab on the right side can be pulled out to reveal aphotograph of the referent against a white background.

Picture books. A 16-page picture book is included with eachvolume of the intervention. Each page includes one target wordprinted on a large flap. The flap can be lifted to reveal a photo-graph of the referent. Fonts vary across words; photographs arepresented in full color on white backgrounds. There is no addi-tional text other than the word on the page.

Research DesignOur research was designed to examine the four phases of read-

ing, from pre-alphabetic to consolidated or conventional reading.Although there is substantial agreement among theories in thephases that distinguish its development, Ehris four-phase model isthe most comprehensive in scope (Ehri, 1994), examining the fulldevelopmental period of learning to read. In her model, each phaseof reading development is characterized by the predominant typeof connection that binds written words to their other identities inmemory: (a) pre-alphabetic, involving visual and contextual con-nections (e.g., using word configurations, or word length withoutany phonological information contributing to the association); (b)

partial alphabetic, making connections between some salient let-ters and sounds (e.g., using the sound values of some letters toform connections between spellings and pronunciation of words);(c) full alphabetic, involving complete connections between graph-emes and phonemes (e.g., able to form connections between allgraphemes in spellings and the phonemes in words, securing thewords in memory); and (d) consolidated alphabetic, when readingbecomes fluent and automatic (e.g., readers can read words as awhole rather than as a sequence of grapheme/phoneme units).

Based on this model, our research design was to examinefeatures of these developmental phases, reasoning that although allchildren in the sample might not become conventional readers(despite market claims), they would likely show evidence of ac-quiring some of the skills critical to its development (see Figure 1).

In the present study, therefore, we used multiple methods ofassessment including parent reports (language development), aswell as 10 eye-tracking tasks throughout the course of 7 months.Eye tracking is a noninvasive methodology permitting high-resolution analyses of eye movement patterns. In addition to over-all visual preference, eye tracking allows a more precise analysisof how infants distribute their attention, such as where infants look(i.e., scanning patterns) and how they shift their gaze from onelocation to another (i.e., saccade latencies; Gredebck, Johnson, &von Hofsten, 2010). Recognizing that young children might ex-hibit implicit knowledge prior to explicit demonstrations, thesegaze-based measures allowed us to tap visual preferences fororthographic knowledge (Golinkoff, Ma, Song, & Hirsh-Pasek,2013) and other related reading skills that might otherwise not berecognized in measures that require overt demonstrations (e.g.,physical actions or verbal responses).

Over the course of 7 months, we conducted a home visit,arranged for each participant to make four laboratory visits toengage in eye-tracking tasks, and performed monthly assessmentsof language development. In addition, we held bi-weekly tele-phone conversations with parents to assess fidelity. Table 2 sum-marizes these measures, which are described in the following text.

Baseline MeasuresTo ensure that our randomization procedures allowed for equiv-

alence between conditions before the intervention, we visitedfamilies in their homes and administered the Bayley Scales ofInfant and Toddler Development (Bayley, 2006), the HOME In-ventory (Caldwell & Bradley, 2003), and a demographic question-naire.

Bayley Scales of Infant and Toddler Development. Infantsgeneral developmental functioning was assessed using the BayleyScales of Infant and Toddler Development III (BSIDIII), a de-velopmental battery yielding norm-referenced scores (Bayley,2006). Three domains were examined during the initial home visit.Infants were administered the Cognitive Scale and the LanguageScale (including both the Receptive and Expressive Communica-tion subtests) by a trained research assistant, while parents com-pleted the SocialEmotional Scale portion of the caregiver ques-tionnaire. The reported reliability for the BSIDIII ranges from .87to .93.

Infant/Toddler HOME inventory. The Infant/Toddler (IT)version of the Home Observation for Measurement of the Envi-ronment (HOME; Caldwell & Bradley, 2003) was administered at

This

docu

men

tis

copy

right

edby

the

Am

eric

anPs

ycho

logi

calA

ssoc

iatio

nor

one

ofi

tsal

lied

publ

isher

s.Th

isar

ticle

isin

tend

edso

lely

fort

hepe

rson

aluse

oft

hein

divi

dual

use

ran

dis

not

tobe

diss

emin

ated

broa

dly.

4 NEUMAN, KAEFER, PINKHAM, AND STROUSE

the initial home visit. The IT HOME uses both observation andinterview to measure the quality and quantity of stimulation andsupport available to infants in their home environment in six areas:(a) Responsivity, (b) Acceptance, (c) Organization, (d) LearningMaterials, (e) Involvement, and (6) Variety. Scores on the sub-

scales were summed (maximum ! 45) to yield an overall HOMEscore.

Demographic questionnaire. Parents were asked to completea 40-item demographic questionnaire. Items included family de-mographic information (e.g., ethnicity, native language), socioeco-

Figure 1. Study outcome measures placed in Ehris (1994) four-phase model of reading.

Table 2Developmental Phases of Learning to Read and Their Associated Measures

Developmental phasea/brief description Measure Adapted from

Pre-alphabeticExpressive vocabulary knowledge MacArthur CDI Fenson et al. (1994)Speed and accuracy of spoken word recognition Speech processing efficiency Fernald et al. (2006)

Partial alphabeticReceptive vocabulary knowledge Vocabulary knowledge Meints et al. (1999); Swingley & Aslin (2000)Phoneme/grapheme correspondence Lettersound knowledge Letter identification (Woodcock, 1987)Knowledge of graphemes Lettername knowledge Word attack (Woodcock, 1987)Identification of written first name Name recognition bKnowledge of standard book format (e.g., upright

orientation) Print awareness (orientation) DeLoache, Uttal, & Pierroutsakos (2000)Knowledge of the properties of letters and words Orthographic knowledge Cassar & Treiman (1997); Kaefer (2009)

Full alphabeticRecognition of previously read words Sight word reading Ehri (1994)Knowledge of words taught in the baby media program Target vocabulary knowledge bKnowledge of standard print format (e.g., right to left,

top to bottom) Directionality of print Clay (1979)Translation of text into words Decoding Word attack (Woodcock, 1987)

Consolidated alphabeticExpressive vocabulary knowledge MacArthur CDI Fenson et al. (1994)Comprehension Reading with meaning b, c

Note. MacArthur CDI ! MacArthurBates Communicative Developmental Inventories.a Adapted from Ehri (1994). b Researcher-developed. c Adapted from promotional materials for the baby media program.

This

docu

men

tis

copy

right

edby

the

Am

eric

anPs

ycho

logi

calA

ssoc

iatio

nor

one

ofi

tsal

lied

publ

isher

s.Th

isar

ticle

isin

tend

edso

lely

fort

hepe

rson

aluse

oft

hein

divi

dual

use

ran

dis

not

tobe

diss

emin

ated

broa

dly.

5RANDOMIZED TRIAL OF BABY MEDIA

nomic background (e.g., parent education, family income), familystructure (e.g., marital status, number of siblings), and educationalactivities (e.g., infants book-reading and baby media experi-ences).

Exit QuestionnaireAt the end of the study, parents were asked to complete a

16-item questionnaire indicating their beliefs about their childrensearly literacy behaviors including their ability to write their name,recognize letters, and read.

Child AssessmentsChildrens first lab visit occurred approximately 2 weeks after

their initial home visit. Follow-up lab visits then occurred afteranother 3 months, 4 months, and 6 months, or after the treatmentgroup finished their initial viewing of Volumes 2, 3, and 5,respectively. At these visits, we examined the four phases ofreading development using a series of eye-tracking tasks.

Eye-tracking technology.Apparatus. Eye movements were measured with a T120 eye

tracker (Tobii Technology, Falls Church, VA) integrated into a17-in. thin film transistor (TFT) monitor (Psychology SoftwareTools, Pittsburgh, PA). This is a remote eye-tracking system thathad no contact with the infant. The typical spatial accuracy of thissystem is approximately 0.5 visual degrees, and the sampling rateis 120 Hz. During tracking, the eye tracker uses infrared diodes togenerate reflection patterns on the corneas of the infants eyes.These reflection patterns, together with other visual informationabout the infant, are collected by image sensors and used tocalculate the three-dimensional position of each eye and gaze pointon screen. The TFT monitor employs active matrix technology inwhich transistors control each pixel on the screen, improvingimage quality and contrast relative to passive-matrix technologies.The monitor has a display resolution of 96 pixels per inch, ensur-ing that images are discernible.

This system uses a binocular tracking method, which allows forincreased head movements. Head movements typically result in atemporary accuracy error of approximately 0.2 visual degrees. Inthe case of particularly fast head movements (i.e., over 25 cm/s),there is a 300-ms recovery period to full tracking ability. Anembedded camera is also used to record infants reactions.

General procedure. Infants sat in a high chair or on theirparents lap approximately 60 cm from the monitor. Parents woreheadphones and blinders to prevent any interference with theirinfants looking behavior. Stimuli were displayed on the Tobiimonitor and a second monitor facing the experimenter. TobiiStudio Professional 3.0 software was used for stimuli presentationand data processing.

To calibrate gaze, an attention grabber was shown at five pointson the screen. A manual calibration procedure was used; accuracywas checked by Tobii Studio software and repeated as necessary.Following calibration, a 2-s attention grabber appeared in thecenter of the screen prior the beginning of each eye-tracking task.During the task, the experimenter monitored infants eye move-ments and behaviors using the live viewer. If infants becamedistracted during the video, the experimenter made a noise (e.g.,snapping or shaking keys) behind the monitor to re-orient their

attention toward the video. Total duration of each eye-tracking taskwas approximately 5 min.

Each visit took approximately 4560 min, including both fa-miliarization and testing. In the case when we used both interactivemeasures, such as the name recognition task and an eye-trackingtask, infants would be given a break between tasks. When therewere two eye-tracking tasks in a single visit, tasks would bepresented consecutively (i.e., no breaks; total testing time approx-imately 10 min). However, breaks were always given if infantswere fussy or if parents requested a break.

Measures were presented in a set order across children. Withineach task, trial order was randomized across children (except forname recognition, which was counterbalanced.) If a child wasnoncompliant on a task, the case would be eliminated for thatparticular measure (approximately 1% of the time). In the casewhen a child was entirely noncompliant, the visit would be re-scheduled. This only occurred twice (with the same family) overthe course of the study.

Children received a small gift, such as a book or toy, at the endof each visit. Parents were compensated $100 for participation inthe study, $50 at the beginning and $50 at its completion. Inaddition, parents received travel expenses (i.e., $0.56/per mile andparking) and were allowed to keep the baby media product at theconclusion of the study.

General data processing. Eye movement data were extractedusing Tobii Studio Professional 3.0 software. Fixations were de-fined as any gaze coordinates lasting at least 60 ms and wereidentified using the Tobii Studio fixation filter. Adjacent gazes(i.e., gazes within a 0.5 radius, lasting less than 75 ms) weremerged into a single fixation.

Assessments at the pre-alphabetic phase.Expressive vocabulary knowledge. Infants expressive vocab-

ulary knowledge was assessed using the short forms of theMacArthurBates Communicative Development Inventories (Fen-son, Pethick, Renda, & Cox, 2000). Parents of infants ages 16months and older were asked to indicate which of the words on theLevel II form their child produced. Parents of infants younger than16 months of age completed the Level I form. To facilitate com-parison across forms, we computed percentile scores for each childbased on his or her age and gender and limited our analysis toexpressive vocabulary.

Speech processing efficiency. Previous research has indicatedthat speed of word recognition during infancy is positively predic-tive of long-term lexical and grammatical development (Fernald,Perfors, & Marchman, 2006). To examine infants speech process-ing efficiency, we had the infants view 12 pairs of referents (i.e.,target and foil) on the eye-tracking monitor. Referents were se-lected from published lexical norms (Dale & Fenson, 1996) asfamiliar to most infants in the age range of our sample. For eachtrial, a pair of 5" 5 in. photographs appeared. After a pre-messagebaseline of 2,150 ms, a voiceover provided a directive (e.g., Lookat the car!). Photographs then remained on the screen for anadditional 2,750 ms. This procedure was then repeated for theremaining trials. Leftright orientation was counterbalanced and pre-sented in a set order; trial order was randomized across infants.Rectangular areas of interest (AOIs) were drawn around each referent.AOIs were kept the same size (461" 457 pixels) for all photographsfor consistency across trials. Fixation duration for each AOI during

This

docu

men

tis

copy

right

edby

the

Am

eric

anPs

ycho

logi

calA

ssoc

iatio

nor

one

ofi

tsal

lied

publ

isher

s.Th

isar

ticle

isin

tend

edso

lely

fort

hepe

rson

aluse

oft

hein

divi

dual

use

ran

dis

not

tobe

diss

emin

ated

broa

dly.

6 NEUMAN, KAEFER, PINKHAM, AND STROUSE

the baseline phases and latency to first fixation on the AOIs during thetest phase were exported for analysis.

Assessments at the partial alphabetic phase.Receptive vocabulary knowledge. To assess infants receptive

knowledge of vocabulary introduced in the baby media program,we created an eight-item measure modeled after Meints, Plunkett,and Harris (1999). For this task, infants viewed pairs of referents(i.e., target and foil) on the eye-tracking monitor. All referentswere featured in the first two volumes of the baby media program.For each trial, a pair of 5 " 5 in. photographs appeared. Imageresolution averaged 466" 520 pixels. To avoid biases due to colorpreference, target/nontarget pairs were also broadly matched forcolor (e.g., a shoe and a hat that were both red).

After a pre-message baseline of 2,150 ms, a voiceover provideda directive (e.g., Look at the ear!). Photographs remained on thescreen for an additional 2,750 ms. This procedure was then re-peated for the remaining trials. Although target words are oftenpresented multiple times in looking-while-listening paradigms(e.g., Fernald, Zangl, Portillo, & Marchman, 2008), we opted forusing a single trial per word to potentially avoid task fatigue.

Leftright orientation was counterbalanced and presented in aset order; trial order was randomized across infants. RectangularAOIs were drawn around each referent; sizes were consistent(461 " 457 pixels) across trials. Fixation duration for each AOIduring the baseline and test phases was exported for analysis.

Lettersound knowledge. To assess infants understanding ofgrapheme-sound correspondences, we created a six-item preferen-tial looking task modeled after the LetterWord Identificationsubtest of the Woodcock Johnson III (Woodcock, McGrew, &Mather, 2001). For each trial, infants viewed a pair of lowercaseletters. After a 2,150-ms pre-message baseline, a voiceover pre-sented a directive (e.g., Look at the /b/!). The procedure wasthen repeated for the remaining trials. Leftright orientation wascounterbalanced and presented in a set order; trial order wasrandomized across infants. Rectangular AOIs were drawn aroundeach referent; sizes were consistent (326 " 312 pixels) acrosstrials. Fixation duration was then exported for each AOI during thebaseline and test phases.

Lettername knowledge. Infants knowledge of graphemename correspondences was assessed through a six-item preferen-tial looking task adapted from the LetterWord Identificationsubtest of the Woodcock Johnson III (Woodcock et al., 2001). Foreach trial, infants viewed a pair of lowercase letters. After a2,150-ms pre-message baseline, a voiceover presented a directive(e.g., Look at the t!). The procedure was then repeated for theremaining trials. Leftright orientation was counterbalanced andpresented in a set order; trial order was randomized across infants.After drawing rectangular AOIs of 326 " 312 pixels around thereferents, we exported fixation duration for each AOI during thebaseline and test phases.

Name recognition. Personal names are one of the first writtenwords recognized by young children (Treiman, Cohen, Mul-queeny, Kessler, & Schechtman, 2007). To examine infants writ-ten name recognition, we created a four-item receptive task. Foreach trial, infants were shown two identical toys (i.e., two greencars or two yellow boats). One toy was labeled with the infantsfirst name in 20-point font against a white background. The othertoy was labeled with a pseudo-word matched in character lengthwith the infants name (e.g., Nathan vs. Gombie). Infants were

shown both toys and asked to select the one bearing their name(e.g., Get Nathans car!). If infants made a selection, theyadvanced to the next trial. If infants failed to make a selection orselected both toys, the trial was repeated. This procedure wasrepeated for a total of two car trials and two boat trials (counter-balanced for toy order and left/right placement). Trials were scoreddichotomously (i.e., correct or incorrect), summed to yield anoverall score, and converted into a proportion score.

Print awareness (orientation). Research suggests that an un-derstanding of the canonical orientation of books and print may beone of the earliest-emerging domains of print awareness (DeLo-ache, Uttal, & Pierroutsakos, 2000). To examine infants under-standing of book and print orientation, we created a six-itempreferential-looking task. For half of the trials, infants viewed pairsof book covers (i.e., upright and inverted); for the remaining trials,they viewed pairs of pseudo-words (i.e., upright and inverted). Foreach trial, infants were oriented to the screen by an attentiongrabber, and then a pair of images appeared. The trial lasted 10 s;there were no oral prompts. This procedure was then repeated forthe remaining trials. Leftright orientation was counterbalancedand presented in a set order; trial order was randomized acrossinfants. For book trials, AOIs of 450 " 503 pixels were drawnaround each cover; for word trials, AOIs of 375" 152 pixels weredrawn over each word. Total fixation duration to each AOI wasthen exported.

Orthographic knowledge. Recent work (Kaefer, 2012) sug-gests that young childrens understanding of orthographic conven-tions may emerge earlier than previously reported. We examinedinfants intuitive orthographic knowledge through a nine-itempreferential-looking task. In three mirror image trials, we paired apseudo-word that obeyed English orthographic conventions withits mirror image. In six illegal character trials, we paired ortho-graphically legal pseudo-words (e.g., pobe) with orthographicallyillegal versions (i.e., p#be). For each trial, infants were oriented tothe screen by an attention grabber, followed by a pair of words.Trials lasted 10 s; there were no oral prompts. Left-right orien-tation was counterbalanced and presented in a set order; trial orderwas randomized across infants. Rectangular AOIs were drawnover each word (438 " 159 pixels), as well as over the individualorthographically illegal characters (106 " 159 pixels). We thenexported total fixation duration to each AOI.

Sight word reading. We examined infants ability to representfamiliar sight words in memory (Ehri & Robbins, 1992) through asix-item preferential-looking assessment. For each trial, infantsviewed a pair of lowercase words (i.e., target and foil) that werefeatured in the baby media program. Following a 2,150-ms pre-message baseline, infants were presented with a directive (e.g.,Look at baby!). Word pairs remained on screen for an additional2,750 ms. The procedure was then repeated for the remainingtrials. Leftright orientation was counterbalanced and presented ina set order; trials were presented in random order across partici-pants. Rectangular 435" 193 pixel AOIs were drawn around eachword, and fixation duration to each AOI during the baseline andtest phases was exported.

Assessments at the full alphabetic phase.Target vocabulary knowledge. Previous research has sug-

gested that young children may acquire little oral vocabularyknowledge from viewing infant-directed DVDs. To assess infantsexpressive knowledge of words introduced in the baby media

This

docu

men

tis

copy

right

edby

the

Am

eric

anPs

ycho

logi

calA

ssoc

iatio

nor

one

ofi

tsal

lied

publ

isher

s.Th

isar

ticle

isin

tend

edso

lely

fort

hepe

rson

aluse

oft

hein

divi

dual

use

ran

dis

not

tobe

diss

emin

ated

broa

dly.

7RANDOMIZED TRIAL OF BABY MEDIA

program, we created a 117-item checklist modeled after theMacArthurBates Communicative Development Inventories (Fen-son et al., 2007). The checklist consisted of all single wordshighlighted in at least one volume of the baby media DVD andincluded in the word cards. Parents were instructed to indicate allwords (including their morphological inflections) currently said bytheir infants. Each checklist was summed to yield an overall scoreand converted into a proportion score.

Directionality of print. Understanding the directionality oftext is a key element of young childrens developing concepts ofprint (Justice & Ezell, 2001). In a nine-item task, we assessedinfants understanding of left-to-right directionality of words (sixtrials) and top-to-bottom directionality of text (three trials). Forword trials, infants viewed a single 27- to 30-character word (i.e.,directionality) or string of symbols (i.e., no directionality). Ifinfants understood the directionality of text, we would expect themto demonstrate a preference for the AOI for real words and nopreference for any of the AOIs for meaningless strings of symbols.For text trials, they viewed several simple sentences taken fromcommercially available childrens books. For each trial, infantswere first oriented to the screen by an attention grabber; the trialthen lasted 10 s. There were no oral prompts. Trial order wasrandomized across infants. For word trials, the width of the mon-itor screen was divided into thirds, and a rectangular AOI (318 "151 pixels) was drawn across each third of the text. For sentencetrials, the screen was divided into quadrants, and AOIs (481" 360pixels) were drawn to cover each quadrant. Tobii Studio Profes-sional 3.0 software was then used to export the location of firstlook for each trial.

Decoding. Conventional literacy is frequently characterized asthe product of code-based skills and comprehension (Gough &Tunmer, 1986). To investigate infants decoding abilities, weconstructed a six-item assessment modeled after the Word Attacksubtest of the Woodcock Johnson III (Woodcock et al., 2001). Foreach trial, infants viewed a pair of words that were not included inthe baby media program (i.e., target and foil). Following a2,150-ms pre-message baseline, infants were presented with adirective (e.g., Look at cheese!). The word pair remained onscreen for an additional 2,750 ms. This procedure was then re-peated for the remaining trials. Leftright orientation was coun-terbalanced and presented in a set order; trials were presented inrandom order across participants. Rectangular 435 " 193 pixelAOIs were drawn around each pseudo-word, and fixation durationto each AOI during the baseline and test phases was exported.

Assessments at the consolidated alphabetic phase.Expressive vocabulary knowledge. Expressive vocabulary

knowledge was assessed using the short form of the MacArthurBates Communicative Development Inventories (Fenson et al.,2000). At the final visit, all infants were age 16 months or older;therefore, all parents were asked to complete the Level II form.Percentile scores were calculated based on infants age and gender.

Reading with meaning. To examine infants ability to com-prehend simple written phrases, we created a six-item task mod-eled after the promotional materials for the baby media program.All phrases were featured in at least one volume of the programDVD and word cards.

To ensure that infants understood the task, we first administeredtwo training trials. The research assistant held up a 5.5 " 8.5 in.white card printed with a target phrase in 72-point lowercase text.

While running her finger left to right under the text, she read thedepicted phrase aloud (e.g., It says shake your head!), orallyrepeated the target action (e.g., Shake your head!), and per-formed the action. This procedure was repeated until the infantalso completed the action.

Following the two training trials, infants completed six test trials(administered in a randomized order). For each trial, the researchassistant held up a card and, while running her finger left to rightunder the text, provided a directive (e.g., Do this one!). If infantsresponded, they were given neutral feedback and moved on to thenext trial. If infants failed to respond, the trial was repeated (up toa total of three repetitions). This procedure was then repeated forthe remaining test trials. Trials were scored dichotomously (i.e.,correct or incorrect) online. Additionally, 20% of video recordingswere randomly selected to be independently coded by a secondresearch assistant. Interrater agreement was 95.83%.

ProcedureFollowing baseline procedures, two trained research assistants

visited treatment families in their homes. They introduced theintervention procedure, adapted from the directions provided bythe baby media program, modeled its use, and answered anyquestions. The same researchers visited all families to ensureconsistency of instruction.

During the home visit, parents were provided with the firstvolume of intervention materials. The research assistant reviewedthe instructions, suggesting that parents show their infants the firstDVD two times per day for 30 days (i.e., 20 hr of exposure). Theywere encouraged to watch the DVD with their babies and point outwords on the screen whenever possible. Parents were also in-structed to engage with their infants while using each of the otherintervention materials (i.e., word cards, picture cards, and picturebook) for 15 min per day (i.e., 7.5 hr of exposure per component)and to feel free to break the interactions up across the day if theyfound their child was unwilling to attend for the full time. Repe-titions of intervention materials were encouraged but not required.Finally, parents were provided with tips for supporting readingactivities every day (e.g., playing matching games, pointing outrhyming words). Approximately 30 days after the home visit,families were mailed the second volume of intervention materials,along with a new set of instructions. Following the programsguidelines, parents were instructed to use the second volume onceper day for 60 days and to follow the previous protocols recom-mendation with each of the other materials. Additionally, parentswere asked to show their children the previous DVD, as well as useeach of the previous materials, once a week.

By following this protocol designed by the program developers,over the course of the study, infants would receive 70 hr of DVDtraining and 45 hr of interacting with each of the other materials(i.e., word cards, picture cards, and picture books). Together, theywould be exposed to 117 words in multiple formats.

Fidelity of implementation. We developed a four-itemchecklist to capture the degree to which treatment families adheredto the instructions for implementing the baby media program.Using the programs parent guide, we identified key features of theprogram to include on the checklist: (a) infant watched the DVD,(b) infant used the word cards, (c) infant used the picture cards,and (d) infant read the picture book.

This

docu

men

tis

copy

right

edby

the

Am

eric

anPs

ycho

logi

calA

ssoc

iatio

nor

one

ofi

tsal

lied

publ

isher

s.Th

isar

ticle

isin

tend

edso

lely

fort

hepe

rson

aluse

oft

hein

divi

dual

use

ran

dis

not

tobe

diss

emin

ated

broa

dly.

8 NEUMAN, KAEFER, PINKHAM, AND STROUSE

Over the course of the study, a trained research assistant calledtreatment families on a bi-weekly basis. Parents were asked abouttheir infants use of the baby media program on the previous day.Checklist items were adapted to specifically target the currentvolume of the program. Each feature was scored dichotomously as1 (i.e., completed) or 0 (i.e., not completed). Implementationvaried across the four components. Families were mostly likely touse the DVD (M ! 64.38%, SD ! 29.56), followed by the picturebook (M! 57.92%, SD! 28.99), and picture cards (M! 54.48%,SD ! 27.64). They were least likely to use the word cards (M !26.90%, SD ! 30.07).

To examine whether enactment declined over the course of thestudy, we compared each familys implementation checklistsacross the five volumes. Fidelity to the full baby media program(all elements together), as well as implementation of the picturebook, picture cards, and word cards, remained consistent acrossvolumes (all ps # .05). However, parents reported significantlylower enactment of the final DVD compared with prior volumes(p ! .014).

Results

In this section, we address the effects of the intervention onbabies reading development. We begin by conducting descriptiveanalyses to examine the distributional properties of the data and todetermine the equivalency of the treatment and control groupsprior to further analysis. Subsequently, we use inferential statisticsto examine the effects of the program. To test whether infantswho used the components of the program more frequently weremore likely to display early literacy skills than children whoused the program less frequently, we correlated fidelity witheach of the outcome measures. There was no evidence thatfidelity to the program supported any of the literacy skills (seeTable 3). Further, we conducted all analyses with percentage offidelity as covariate and found it to be nonsignificant for allmeasures. Therefore, we used one-way analyses of covariance(ANCOVA) with condition as the independent variable, and base-line as covariate when included in the task, to examine each phaseof development in learning to read. Because of the age rangeamong the children and the developmental differences betweeninfants at various stages, we included age as a covariate in allnonstandardized analyses. Additionally, when available, we usedpretest or baseline scores as covariates.

Descriptive AnalysesAs shown in Table 1, there were no significant group differences

in the child demographic data, HOME scores, or Bayley scores atthe infants first visits.

Treatment and control groups also did not differ initially on theirprevious media experience. Data for our sample are presentedalongside national averages in Table 4. In general, children in oursample had less television exposure than the national average, hadfewer televisions in their homes, and were less likely to have cableaccess than the average child (Rideout & Hammel, 2006).

Contrastingly, infants in our sample in both treatment and con-trol groups were more likely to have shared-reading experiencethan the national average, with 75% of the parents reporting theyread daily to their child (see Table 5). Despite having more regular

reading sessions than the national average, infants in our sampleappeared to be read to for slightly less duration than the averageinfant, with the most popular response for parents in our samplebeing 1530 minutes, just below the national average of 33 min-utes per day.

Pre-Alphabetic and Partial Alphabetic PhasesMeans and standard deviations for the pre-alphabetic and partial

alphabetic phases of reading are presented in Table 6. As shown,although the means were slightly higher for the treatment group inthe pre-alphabetic phase, there were no significant differencesbetween groups on either of these measures. As evidenced byparental report, expressive vocabulary between groups was statis-tically equivalent.

Speech processing efficiency. Similarly, there were no sig-nificant differences between groups in childrens ability to processspeech. If the treatment had facilitated childrens processing oforal language, we would have expected children to demonstratesignificantly lower time to first fixation than children in the controlgroup. To measure such processing, we first checked to make surethat children did not have a preference for the target or foil pictureprior to being prompted to look. In our task, seven target wordspassed this first criterion: bottle, bucket, car, dog, ear, giraffe, andtiger. Second, to ensure that childrens latency to fixate was basedon actual processing of the verbal prompt, we needed to establishthat children knew the words presented in the task. We confirmedthis by asking parents to indicate which of the words their childrenunderstood and excluded trials for which we had no parent con-firmation. Each childs latency to look to the target was thenaveraged across each of their eligible trials. There was no differ-ence between groups in latency to look at the target object, F(1,75) ! 1.77, p ! .188, &2 ! .023. The age covariate was alsononsignificant, F(1, 75)! 1.33, p! .258, &2! .017. These resultssuggest that after more than a months intervention, there was noevidence of significant effects for treatment on childrens pre-alphabetic skills in the ability to process speech more efficiently.

Table 3Correlation Between Fidelity to the Intervention andChild Outcomes

Outcome measure r p

Expressive vocabulary knowledge .255 .056Receptive vocabulary knowledge '.053 .710Lettersound knowledge '.105 .465Lettername knowledge .048 .743Name recognition .200 .168Print awarenessbook orientation .075 .600Print awarenesstext orientation '.152 .290Orthographic knowledgemirror .102 .482Orthographic knowledgeillegal character .108 .456Sight word reading '.283 .045Target vocabulary knowledge .227 .102Directionalitywords .039 .786Directionalitysentences .170 .233Decoding .132 .356Expressive vocabulary knowledge .125 .371Reading with meaning '.024 .863

Note. None of the measures had ps greater than the Bonferroni-correctedvalue of .003.

This

docu

men

tis

copy

right

edby

the

Am

eric

anPs

ycho

logi

calA

ssoc

iatio

nor

one

ofi

tsal

lied

publ

isher

s.Th

isar

ticle

isin

tend

edso

lely

fort

hepe

rson

aluse

oft

hein

divi

dual

use

ran

dis

not

tobe

diss

emin

ated

broa

dly.

9RANDOMIZED TRIAL OF BABY MEDIA

At the partial alphabetic phase, we found a similar pattern.Given the number of tasks at this phase, we would expect that theintervention might influence at least some of the initial speech-to-print skills that it presumably emphasizes throughout the program.However, as shown in Table 5, there were no apparent patterns ofimprovements in these skills, with the treatment group slightlyahead on four of the measures, and the control group slightly aheadon another four of the measures. None were statistically signifi-cant, as described in the following text.

Receptive vocabulary knowledge. If children learned thevocabulary words presented in the baby media program, we wouldexpect children in the treatment group to recognize more of thetarget vocabulary words taken from the program than those in thecontrol group. A one-way ANCOVA indicated that there was no

difference between groups in proportion of time looking to thetarget photographs at baseline, F(1, 101) ! 0.56, p ! .458, &2 !.005. The age covariate was also nonsignificant, F(1, 101) ! 0.06,p! .805, &2! .001. There was also no difference between groupsin the proportion of time looking to the target photographs afterthey were prompted, F(1, 102) ! 1.32, p ! .254, &2 ! .013. Ageand baseline fixations were both nonsignificant covariatesage:F(1, 102) ! 0.61, p ! .439, &2 ! .006; baseline fixations: F(1,102) ! 3.35, p ! .070, &2 ! .033.

Lettername and lettersound knowledge. If exposure tothe treatment enhanced letter knowledge, we would expect chil-dren in this condition to demonstrate significantly longer fixationsto the target letter (compared with the foil letter) than children inthe control group. One-way ANCOVAs with condition as the

Table 4Infants Television Exposure

Questionnaire itemTreatment(n ! 61)

Control(n ! 56) pa

Nationalaverage

No. of working televisions in the home (%)0 3.3 6.0 .343 1%b13 90.0 91.0 75%b4 or more 6.7 5.4 24%b

Home has cable/satellite television 63.3 69.6 .472 80%bInfant has started watching television 56.7 42.9 .137 79%cInfants average daily television viewing 34 minc

None 64.4 60.7 .446$ 1 hr 25.4 33.912 hr 10.2 5.4

Parent talks with infant while co-viewingNever 14.3 20.5 .177 Once in a while 16.3 10.3Frequently 53.1 38.5Almost always 16.3 30.8

Infant has television in bedroom 3.3 5.5 .577 19%cInfant has favorite TV program/DVD 41.7 28.6 .140 Parent attitudes toward educational mediad 2.67 (0.43) 2.34 (0.61) .002!!

Mostly helpsb 38%Not much effectb 22%Mostly hurtsb 31%

a p reported for t test or chi-square test of treatment versus control groups. b Data reported by Rideout andHammel (2006) for children ages 6 months6-years. c Data reported by Rideout and Hammel (2006) forchildren ages 6-months23-months. d Mean value and standard deviation on scale of 14; neutral score is 2.5.!! p $ .01.

Table 5Infants Shared Reading Experience

Questionnaire item

Percentage of those in

paNationalaverageb

Treatment group(n ! 61)

Control group(n ! 56)

Infant has started being read to 98.4 100 .332 94%cInfant is read to daily 71.7 78.6 .321 58%cInfants average daily reading duration 33 min

Not at all 1.7 0.0 .485A few minutes 25.0 39.3At least 15 min 55.0 46.4More than 15 min 18.3 14.3

Infant has favorite book 63.3 58.9 .627 n/aa p reported for chi-square test of treatment versus control groups. b Rideout and Hammel (2006) for childrenages 6-months23-months.

This

docu

men

tis

copy

right

edby

the

Am

eric

anPs

ycho

logi

calA

ssoc

iatio

nor

one

ofi

tsal

lied

publ

isher

s.Th

isar

ticle

isin

tend

edso

lely

fort

hepe

rson

aluse

oft

hein

divi

dual

use

ran

dis

not

tobe

diss

emin

ated

broa

dly.

10 NEUMAN, KAEFER, PINKHAM, AND STROUSE

independent variable and age as the covariate indicated that therewas no difference between groups in proportion of time looking tothe target letter at baseline for letter sounds, F(1, 100)! 1.37, p!.245, &2! .014, or letter names, F(1, 100)! 1.12, p! .293, &2!.011. The age covariate was also nonsignificant in both testssounds: F(1, 100) ! 0.02, p ! .891, &2 $ .001; names: F(1, 100)0.55, p ! .462, &2 ! .006. This assessment demonstrated thatchildren had no visual preference for one of the letters over theother prior to being prompted where to look.

One-way ANCOVAs with age and baseline looking as covari-ates indicated that there was also no difference between groups inproportion of time looking at the target letter at test (after theprompt) when the letter sound was prompted, F(1, 99) ! 0.002,p ! .968, &2 $ .001. Both covariates were nonsignificant aswellage: F(1, 99) ! 0.006, p ! .937, &2 $ .001; baseline: F(1,99) ! 0.63, p ! .430, &2 ! .006. Additionally, there was nodifference between groups in proportion of time looking at thetarget letter when the letter name was prompted, F(1, 95) ! 1.67,p ! .200, &2 ! .015. Age and baseline looking were both signif-icant covariatesage: F(1, 95) ! 4.14, p ! .045, &2 ! .038;baseline: F(1, 95) ! 8.60, p ! .004, &2 ! .078. Both groupsproportion looking to the target was near chance (.50) for letternames and sounds.

Name recognition. To test whether experience with the inter-vention improved childrens ability to recognize their writtennames, we analyzed childrens performance on four trials in whichthey were asked to select the toy with their name printed on it.Children received a proportion score for the number of trials inwhich they chose the correct object out of four. Both groups scorednear chance (.50). An ANCOVA with age as the covariate indi-cated that there was no difference between the treatment and

control groups, F(1, 98) ! 1.55, p ! .216, &2 ! .015. Age wasalso nonsignificant, F(1, 98) ! 2.23, p ! .139, p ! .022.

Print awareness. If experience with the intervention sup-ported childrens understanding of print orientation, we wouldexpect children in the treatment condition to look longer at uprightbook covers and words than children in the control group. How-ever, there was no group difference in proportion of looking to theupright book covers (vs. the inverted ones), controlling for age,F(1, 95) ! 0.21, p ! .645, &2 ! .002. Age was a nonsignificantcovariate, F(1, 95)! 0.045, p! .833, &2! .001. There was alsono difference in proportion of looking to the upright words, con-trolling for age, F(1, 90) ! 0.04, p ! .837, &2 ! .001. Age wasagain nonsignificant, F(1, 90) ! 1.12, p ! .294, &2 ! .012. Bothgroups spent approximately half the time looking at the target andhalf the time looking at the foil on both tasks.

Orthographic knowledge. Similarly, we would expect chil-dren in the intervention to begin to recognize what was or was nota word. We compared childrens proportion looking to twodifferent types of standard and nonstandard pseudo-words. First,we looked at childrens preference for standard versus mirror-image-reversed copies of the same word. An ANCOVA with ageas the covariate indicated that there was no group difference inproportion looking to the standard word over the reversed word,F(1, 93) ! 0.32, p ! .573, &2 ! .003. Age was nonsignificant,F(1, 93) ! 0.07, p ! .794, &2 ! .001. Next, we investigatedchildrens preference for standard words versus those with illegalcharacters inserted (such as # or $). Again, there were no groupdifferences, controlling for age, F(1, 94) ! 1.64, p ! .203, &2 !.017. Age was nonsignificant, F(1, 94) ! 0.64, p ! .426, &2 !.007. However, we did note that children in the treatment groupspent marginally more time (in raw seconds) fixated on the indi-

Table 6Descriptive Statistics for Pre-Alphabetic and Partial Alphabetic Measures

MeasureTreatment

M (SD)ControlM (SD) pa

Expressive vocabulary knowledge (Percentile score on the MacArthur CDI short form at home visit) 35.81 (30.17) 31.58 (29.33) .444Speech processing efficiency (Latency to look to target in seconds) 0.99 (0.68) 0.80 (0.65) .188Receptive vocabulary knowledge (Proportion of time spent looking to target)

Baseline .57 (.10) .56 (.11) .458Test .59 (.12) .55 (.14) .254

Lettersound knowledge (Proportion of time spent looking to target)Baseline .46 (.14) .49 (.14) .245Test .50 (.16) .50 (.19) .968

Lettername knowledge (Proportion of time spent looking to target)Baseline .48 (.15) .51 (.15) .293Test .48 (.19) .44 (.22) .200

Name recognition (Proportion of trials correct) .53 (.19) .59 (.21) .216Print awareness (Orientation; proportion of times spent looking to target)

Books .54 (.14) .56 (.12) .645Text .47 (.24) .47 (.27) .837

Orthographic knowledgeProportion of time spent looking to target words)

Mirror .53 (.23) .52 (.23) .573Illegal character .38 (.15) .42 (.17) .203

Looking time to illegal character (in seconds) 0.68 (0.81) 0.46 (0.38) .071Sight word reading (Proportion of time spent looking to target)

Baseline .49 (.17) .51 (.17) .666Test .52 (.23) .50 (.22) .413

Note. MacArthur CDI short form ! MacArthurBates Communicative Development Inventoriesshort form (Fenson et al. (2004).a p reported for analysis of covariance comparison of condition.

This

docu

men

tis

copy

right

edby

the

Am

eric

anPs

ycho

logi

calA

ssoc

iatio

nor

one

ofi

tsal

lied

publ

isher

s.Th

isar

ticle

isin

tend

edso

lely

fort

hepe

rson

aluse

oft

hein

divi

dual

use

ran

dis

not

tobe

diss

emin

ated

broa

dly.

11RANDOMIZED TRIAL OF BABY MEDIA

vidual illegal character than children in the control group, control-ling for age, F(1, 98)! 3.32, p! .071, &2! .033, indicating somerecognition of irregularity. Age was nonsignificant, F(1, 99) !0.01, p ! .937, &2 $ .001.

Sight word reading. Finally, if children learned wordtextmappings from exposure to the intervention, we would expect tochildren in the treatment group to demonstrate significantly longerfixations to target words than foil words when prompted. AnANCOVA with age as the covariate indicated that there was nogroup difference in proportion of looking to the target (vs. the foil)prior to prompting, F(1, 100) ! 0.19, p ! .666, &2 ! .002. Bothgroups spent about half the time looking to each side of the screen.These results indicated that children did not prefer to look at onepicture over the other prior to being prompted where to focus.

Baseline looking was significant, F(1, 98) ! 6.14, p ! .015,&2 ! .059. However, a one-way ANCOVA with age and baselinelooking as covariates reported that there were no group differencesin proportion of looking to the target after prompting, F(1, 98) !0.68, p ! .413, &2 ! .007. Age was a nonsignificant covariate,F(1, 98) ! 0.00, p ! .959, &2 $ .001.

In sum, these results showed no evidence of positive effects ofthe intervention on childrens pre-alphabetic or partial alphabeticphases of early literacy development. Using multiple measuresdesigned to tap many different aspects of early development, wefound no discernable significant differences between groups onword learning or the skills associated with reading development.

Full Alphabetic and Consolidated Alphabetic PhasesAlthough one might assume that the latter phases of reading are

predicated on improvements in the earlier phases, and not likely toshow evidence of impact on childrens reading development, webelieved it was prudent to examine the skills associated withconventional reading for several reasons. First, there was a rea-

sonable expectation that reading development may not represent aprocess where one skill is prerequisite for movement to the nextphase (Ehri & Roberts, 2006). None of the developmental theoriesof reading are so rigid. Second, although the instructional design ofthe program is based on an analytic phonics approach, it focusesmostly on sight word reading and associative learning with wordsand word meaning connections; and third, claims made by theseprograms argue for conventional and fluent reading. Therefore, inthe final months, we examined the more conventional skills asso-ciated with the simple view (e.g., decoding and comprehension).Table 7 presents the means and standard deviations of measuresthat are representative of the transition to the consolidated phaseand conventional reading.

Directionality of print. Conceivably, the intervention shouldhelp children understand concepts of print, particularly the direc-tionality of text. Therefore, we would expect children in thetreatment group to direct their gaze toward the beginning of wordsand sentences significantly more frequently than children in thecontrol group. We counted the number of trials on which eachchilds first look was to the upper left portion of the text.Because the data were very heavily skewed (most children didnot look left), we opted to use a KruskalWallis test instead ofan ANCOVA. The KruskalWallis test is the nonparametric al-ternative to the analysis of variance that is used when data arenonnormal (Rosner, 2011). It does not allow for the addition of acovariate, so we first checked for a correlation between age and thenumber of times childrens first look was on the left portion of text.There was no correlation between age and number of left looks forword slides or sentence slides. The KruskalWallis test indicatedthat there were no group differences in the number of left looks forwords (p ! .652) or sentences (p ! .838).

Decoding. If children learned wordtext mappings from YourBaby Can Read, we would expect children in the treatment group

Table 7Descriptive Statistics for Full Alphabetic and Consolidated Alphabetic Measures

MeasureTreatment

M (SD)ControlM (SD) pa

Target vocabulary knowledge (Proportion of target wordschild says) .58 (.33) .41 (.32) $.001!!!

Directionality of print (First look to correct position)Words

Mean number .27 (.67) .24 (.60) .652Range (max ! 6) 04 correct 03 correct

SentencesMean number .59 (.83) .57 (.72) .838Range (max ! 3) 03 correct 02 correct

Decoding (Proportion of time spent looking to target)Baseline .50 (.16) .54 (.21) .357Test .47 (.20) .49 (.24) .744

Expressive vocabulary knowledge (Percentile score on theMacArthur CDI short form at final visit) 46.8 (30.15) 40.3 (32.2) .289

Reading with meaning (No. of behaviors performed)Familiar cues 4 0 .064Novel cues 2 1 .507

Note. MacArthur CDI short form ! MacArthurBates Communicative Development Inventoriesshort form(Fenson et al. (2004).a p reported for analysis of covariance, KruskalWallis, or Fischers exact comparisons of condition.!!! p $ .001, after controlling for age.

This

docu

men

tis

copy

right

edby

the

Am

eric

anPs

ycho

logi

calA

ssoc

iatio

nor

one

ofi

tsal

lied

publ

isher

s.Th

isar

ticle

isin

tend

edso

lely

fort

hepe

rson

aluse

oft

hein

divi

dual

use

ran

dis

not

tobe

diss

emin

ated

broa

dly.

12 NEUMAN, KAEFER, PINKHAM, AND STROUSE

to fixate significantly more on the target words than children in thecontrol group. A one-way ANCOVA, controlling for age, indi-cated no difference in childrens preference for the target over thefoil in the baseline phase (prior to being prompted where to look),F(1, 100) ! 0.86, p ! .357, &2 ! .008. Age was also nonsignif-icant, F(1, 100)! 0.17, p! .679, &2! .002. This established thatchildren had no visual preference for one object over the other.Baseline looking proportion was then entered along with age as acovariate in a one-way ANCOVA to test for group differences inlooking to the target during the test phase (after the prompt wasgiven). There was no group difference, F(1, 97) ! 0.11, p ! .744,&2 ! .001. Age was a nonsignificant covariate, F(1, 97) ! 0.41,p ! .523, &2 ! .004. Baseline looking was significant, F(1, 97) !4.70, p ! .033, &2 ! .046. This indicated that childrens visualpreference in the baseline was the best predictor of their lookingduring test.

Program vocabulary. Toward the end of the study, we askedparents in both groups to identify target words that their childcould say, using a similar format as the MacArthurBates Com-municative Development Inventories, only with target words di-rectly from the intervention program. Comparing treatment andcontrol groups, there was a significant main effect of condition onchildrens expressive knowledge of words introduced in the pro-gram, F(1, 102) ! 5.99, p ! .016, &2 ! .055, after controlling forage. According to parent reports, children in the treatment groupknew an average of 58% of the target words (SD ! 33%), and thecontrol group knew only 41% (SD ! 32%). Age was a significantcovariate, F(1, 102) ! 39.46, p $ .001, &2 ! .279.

Vocabulary knowledge. At the same time, we administeredthe MacArthurBates Communicative Development Inventories toboth groups in this final phase of the study. Because the percentileranking is based partially on childrens age, age was not used as acovariate in the analysis. Although children in the treatment groupwere reported to know more words as indicated by the higher meanpercentile ranking (M ! 46.8, SD ! 30.15) than children in thecontrol group (M ! 40.3, SD ! 32.2), this difference was non-significant, F(1, 104) ! 1.14, p ! .289, &2 ! .011. Therefore, ifchildren in the treatment group were using more expressive vo-cabulary than those in the control group, it was likely due to theirrepeating the words that they heard and saw in the program and notdue to a significant increase in vocabulary knowledge at large.

Reading with meaning. In our final analysis, we examinedchildrens ability to read with meaning using written cue cards.Three of these cue cards contained exact phrases and actionstaught in the program (e.g., Clap your hands), and the other threecards contained words used in the program but combined in novelways (e.g., Touch your face). Childrens responses were video-taped and coded by the experimenter for whether the child pro-duced the requested action.

There were a total of four positive responses to the familiar cues,all of which were produced by the treatment group. We performeda Fischers exact test to determine whether the proportion ofresponses in the treatment group was significantly greater than thenull performance of the control group. We used a one-tailed testdue to the nature of the directional hypothesis (i.e., babies beingable to read), suspecting that it would be rare and that the opposite(i.e., babies failing to respond) might be more common. In otherwords, we did not expect an extreme contingency table indicatingcommon positive (reading) responses. The one-tailed Fischers

exact significance was p ! .064, indicating that there was anonsignificant association between positive responses to the pro-gram cue cards and condition. For the novel cues, there were onlya total of three positive responses, two infants in the treatmentgroup and the other in the control group. The one-tailed Fischersexact significance was p! .746, indicating no association betweenpositive responses and treatment group.

We then asked each of the children who had performed at leastone of the reading behaviors to return to the lab after 6 months;five of the seven children returned for a visit. We repeated thesame set of procedures and prompts as in the initial reading withmeaning task. None of the children successfully completed any ofthe behavioral responses on the written cue cards.

Finally, we examined the exit questionnaire responses of parentsof children who had appeared to read cards on the task versus therest of the sample on two exit items: My child has started to learnto read, and My child knows how to read. KruskalWallis testsindicated that these parents had given their children higher ratingson both items, ps ! .021 and .005, respectively.

In sum, following the use of the intervention over a 7-monthperiod, there was no evidence to indicate that babies in the treat-ment group could read or attend to words and texts any differentlythan children in the control group. Although parents in the treat-ment group reportedly indicated that children knew significantlymore target words in the program than those in the control group,these gains were not evident in the standardized vocabulary mea-sure. Finally, those children who appeared to read and respond towritten language cues seen on the DVDs were not able to identifywords or phrases after the intervention was completed, despite thefact that their parents believed that they were beginning to read.

DiscussionThe last decade has seen the explosion of baby media targeted

to promoting infants development (Rideout & Hamel, 2006).Among the best-selling products are those that claim to teachbabies to read. In their claims, program developers have arguedthat by accelerating the reading process, their products can enableyoung children to advance more rapidly in school, reading com-plex texts which would otherwise be taught in later grades. Thesedevelopers implicitly make the case that by gaining more knowl-edge through text, a child can begin to read earlier and that theearlier a child can begin to read, the more proficient he or she willbecome in school and beyond.

To our knowledge, this is the first study to fully test theseclaims. Purposely, we set out to be most scrupulous in our defi-nition of reading, examining outcomes that not only includedconventional reading but ones that could tap the earlier precursorsof literacy skills. To avoid the limitations of previous studies, wefollowed the program developers detailed protocol, trained par-ents in how to use these products, and conducted ongoing fidelitychecks to measure and assess dosage. We used outcome measuresthat included parent reports and a series of eye-tracking tasks thatcould more sensitively gauge the subtleties of early orthographicand phonological knowledge. Finally, we conducted what is re-garded as the gold standard in research design, randomly assign-ing children and their families to treatment and control groups.

Our results indicated that babies did not learn to read. In total,out of 14 different measures of early reading skills, there were 13

This

docu

men

tis

copy

right

edby

the

Am

eric

anPs

ycho

logi

calA

ssoc

iatio

nor

one

ofi

tsal

lied

publ

isher

s.Th

isar

ticle

isin

tend

edso

lely

fort

hepe

rson

aluse

oft

hein

divi

dual

use

ran

dis

not

tobe

diss

emin

ated

broa

dly.

13RANDOMIZED TRIAL OF BABY MEDIA

null findings. We saw no evidence for the effects on conventionalreading, as program developers had indicated on their promotionalwebsites and testimonials, or on any of the pre-alphabetic or partialalphabetic phases of reading. Even with a greater dosage of treat-ment than in previous studies, there were no effects of the inter-vention on childrens speech processing efficiency, word learningskills, phonological processing, orthographic knowledge, letterrecognition, sight word reading, or reading with meaning.

Nevertheless, based on our exit interviews and parent reportsof target word learning (i.e., words seen on the DVD), there wasthe belief among parents that their babies were learning to readand that their children had benefited from the program in theirexpressive vocabulary development. Parents in the interventionreported that their infants used more of the targeted words fromthe program than those in the control group. This did notgeneralize, however, to the standardized measure of expressiveword knowledge. In this case, there were no reported differ-ences between groups.

These results suggest that parents may have interpreted imita-tion and mimicking as an indicator of word learning. A plethora ofstudies (e.g., Banduras classic bobo doll experiment; Bandura,1965; Neuman, 1991), for example, have shown childrens abilityto mimic what they see on the screen. This mimicking phenome-non was clearly evident in several of the childrens responses towritten commands on cue cards. Although four children initiallyresponded to a phrase prompt directly from the program immedi-ately after the intervention, none were successful a few monthslater. Further, none of the children could respond to words in novelphrases. Consequently, testimonials and videos of reading onmany of these websites may be preying on parents wishes andbeliefs for their childs precocity and not on the reality of whatconstitutes meaningful learning. Although we cannot say with fullassurance that infants at this age cannot learn printed words, wecan confidently say that they did not learn printed words from aproduct of this nature.

Our findings provide further support for experimental studiesthat have demonstrated a lack of significant effects of baby mediaon receptive language. In two previous studies, for example, nei-ther Robb et al. (2009) nor DeLoache et al. (2010) reported asignificant relationship between DVD viewing and infant receptivevocabulary. These results stand in contrast to those of Vandewater(2011) who reported significant positive effects on receptive vo-cabulary growth using a similar program. Comparing her resultswith these previous studies, Vandewater argued that the differ-ences in these findings could potentially be attributed to samplesize or a sleeper effect in which effects were found 3 months afterthe intervention. However, in our study, children were likely to beexposed to a far greater dosage than in Vandewaters study (i.e.,children in her study saw the video an average of 14 times,whereas ours were instructed to view the initial DVD at least 60times, with further redundancy built into all of the program mate-rials and repeated throughout the entire five-volume program) withno evident immediate or sleeper effects. Given that Vanderwatersresults relied on parental report of childrens language gains andnot direct assessments with children, such findings might reflect asocial desirability bias or wishful thinking on the part of parents.DeLoache et al. (2010), who reported both parent opinion andword learning directly assessed with the child, found that par-ents who enjoyed the DVD overestimated how many words

their children learned from it. Although there is some researchto suggest that young children are capable of learning individ-ual words from screen media sources like video (Allen &Scofield, 2010; Krcmar, 2010; Krcmar et al., 2007), it may bea poor substitute for language experiences with live speakers(Golinkoff & Hirsh-Pasek, 1999).

At the same time, we did not find evidence for suppression oflanguage scores, as reported in surveys and correlational stud-ies. Zimmerman, Christakis, and Meltzoff, (2007), for example,reported that watching baby videos predicted significant decre-ments in vocabulary size for infants between 8 and 16 months.In contrast, we found no declines in vocabulary for either groupthroughout our study. Similarly, Linebarger and Vaala (2010)have proposed that the expository content in baby videos (i.e.,such as the program intervention in this study), compared withnarrative or story-like content, may impair language develop-ment due to the volume of information presented in theseprograms. Once again, we found no evidence of decline inlanguage or any other skill as a result of the intervention.Finally, media researchers have suggested that certain featureslike hearing verbal labels for objects visually depicted on-screen, verbal and visual emphasis on novel words and theirreferents, and repetition of verbalizations might support in-creased language acquisition (Krcmar, 2010; Naigles & May-eux, 2001). However, even with repeated exposure (i.e., 20words " 30 days " 4 types of media), we found little supportfor language acquisition or skill development from baby media.

Rather, an alternative hypothesis is that babies are neitherhelped nor harmed by baby media. Clearly, infants can make senseof some video displays. For example, we could not have conductedour eye-tracking studies without some sustained attention to screenmedia. Moreover, studies comparing video instruction with noinstruction at all have consistently shown that infants do learnsome information from videos (Barr & Hayne, 1999; Strouse &Troseth, 2008). Further, there are a number of impressive studiessuggesting that older infants and toddlers, given the right type ofexperiences, can bring their perceptual skills and general knowl-edge backgrounds to bear on a novel problem-solving tasks pre-sented through screen media (Nielsen, Simcock, & Jenkins, 2008;Troseth, Saylor, & Archer, 2006). Nevertheless, the absence of anysignificant effects on childrens language and skills suggests to usthat the developmentalists are most accurate in their recognition ofinfant capabilities and limitations. Infants do bring a limited set ofexperiences and little background knowledge of the content andformat used to deliver information on screen. As a result, they arein poor position to use screen media as a tool for learning how toread, given that language development and background knowledgeare required for reading performance even at the initial level(Neuman, 2006).

Ultimately, therefore, it is about choice. Parents must weighwhether such exposure to media is displacing other activities(Neuman, 1991), such as adultchild language interaction, readingbooks, play, and joint activity. These are the activities, shownthrough a large convergence of research (Neuman & Celano, 2012;Snow, Burns, & Griffin, 1998), that have strong empirical supporton childrens affect, cognitive development, early reading skills,and, in the long run, reading performance.

This

docu

men

tis

copy

right

edby

the

Am

eric

anPs

ycho

logi

calA

ssoc

iatio

nor

one

ofi

tsal

lied

publ

isher

s.Th

isar

ticle

isin

tend

edso

lely

fort

hepe

rson

aluse

oft