Poster group
Details of individual items:
poster
For infants to comprehend the meanings of spoken words they must beable to store and retrieve the sound patterns of words. Recentresearch reveals that this ability emerges at around 4-8 months of age(Jusczyk & Aslin, 1995; Jusczyk & Hohne, 1997; Mandel, Jusczyk, &Pisoni, 1995; Saffran, Aslin, & Newport, 1996). In addition, infantsmust also link these sound patterns to referents in the world in orderto comprehend word meanings. Tincoff & Jusczyk (1999) demonstratedthat 6-month-olds can link the sound patterns, 'Mommy' and 'Daddy' tovideo images of the appropriate parents, but do not attach these wordsto unfamiliar men and women. Their study raised the possibility thatinfants' earliest word-world associations might all be highly specific,as is the case with proper nouns. In the present investigation we test the proposal that infants' earlyword-world associations are highly specific. Whereas proper nouns,such as 'Mommy' do denote specific individuals, common nouns, such as'feet', refer to categories of objects, not just specific individuals. Infants might also initially attach sound patterns of common nouns tospecific and familiar referents, and therefore might not evidencecomprehension if a different and unfamiliar exemplar was presented asthe referent. However, 3- to 4-month-olds can form categories such as'cat', 'mammal', or 'animate' based on perceptual similarities (Quinn &Eimas, 1998). Therefore, 6-month-olds' might be able to link soundpatterns of spoken words to categories of perceptually similarobjects. 6-month-olds viewed videotapes of an unfamiliar female adult's handand feet in silence and while listening to repetitions of 'hand' and'feet'. These items were chosen based on parental reports of word usewith 6-month-olds and because of the high amount of sensory experienceinfants have with their own hands and feet. We calculated a lookingindex score for each infant. This measure adjusts for individualpreferences for the videos, (Test LT Video 1 - Baseline LT Video 1) /(Baseline LT Video 1). To determine if 6-month-olds looked longer atthe video being named, a repeated measures ANOVA with the withinsubjects factors of Named (named vs. unnamed video), Test Item (handvs. feet video), and Test Half (first vs. second test half) wasconducted. There was a significant interaction of Named x Video [F(1,24) 5.023, p .035] and a main effect of Named [F(1,24) 5.482, p .028]. Infants looked reliably longer at the named video for both'hand' and 'feet', though the difference was greater for the 'feet'stimuli (mean difference for hand +9%, for feet +22%). Overall, 19of the 25 infants looked longer at the named video (p .015). Therefore, 6-month-old infants demonstrate some ability to attachsound patterns of spoken words to new and unfamiliar referents that areperceptually similar to familiar referents. We are currentlyinvestigating the influences of amount of experience and movement on6-month-olds' comprehension of common nouns.
poster
no abstract
poster
Soon after birth infants from monolingual families are able todiscriminate utterances drawn from languages that differ prosodically,but being able to perceive a difference between languages that belongto the same rhythmic class seems to take longer. Recent researchindicates that, by two months, languages from the same rhythmic class,such as English and Dutch, are still not discriminated by two months,and it is not till the age of four months that a within-classdiscrimination has been observed for a pair of Romance languages, suchas Catalan and Spanish. Early discrimination capacities are, thus,initially constrained by the rhythmic properties of the languages thatare being compared, but after a certain amount of exposure to aspecific language, infants' perceptual abilities have somewhat beenrefined and make them able to reach a within-class distinction. Inorder to further explore the constraints on these early within-classdiscrimination abilities and at the same time to identify somerelevant dimensions in the speech signal that can account for thesemore fine-grained perceptual abilities, a series of experiments wererun with four-month-old infants growing up in either Spanish orCatalan monolingual families. The language selected for the differentcomparisons was Italian, another Romance language from the samesyllable-timed category, which shares several phonological featureswith both Spanish and Catalan, but which also differs from them inseveral dimensions (i.e. it has a vowel inventory closer to Catalan,but the absence of vowel reduction approximates it to Spanish). ASpanish-Italian comparison was first run, using a visual orientationlatency measure. Materials were 14 utterances in each languagerecorded from a multilingual female speaker. The dependent variablewas the time the infant needed to initiate a saccade towards one oftwo lateral loudspeakers from where the utterances were randomlypresented. Twenty infants, half from Spanish and half from Catalanmonolingual environments participated in the study. Results indicatedthat discrimination was not reached, as neither the Spanish group northe Catalan one showed a differential response. Another twenty infantswere then tested with the Catalan-Italian comparison. This time,results indicated that discrimination was possible, with a clearinteraction between the language of the utterances and the language ofexposure, that is, Spanish infants oriented faster towards Italianutterances while Catalan infants oriented faster towards theutterances in their maternal language. The pattern of results seems tosuggest that Spanish and Italian are difficult to tell apart, evenwhen Spanish has been the language of exposure, while Catalan seemseasier to differentiate from Italian (in a similar way as it had beendistinguished from Spanish in previous research). Moreover, Spanishinfants seem to treat Italian utterances as if the belonged to theirnative language. A possible interpretation of this pattern of resultsis given in terms of the vocalic systems of the languages at issue,and the role of vowel reduction, both from a rhythmic and adistributional perspective, is analysed.
poster
Infant studies of cross-language consonant perception have shown ashift from a language-general to a language-specific pattern during thefirst year of life. However, previous studies have examined only sets ofsounds from a few phonetic categories (e.g. stop consonants, semivowels).The present study expands these findings to a new set of phoneticcategories, fricative vs. affricate, to assess language-specificinfluences on consonant perception by infants. In a previous study (Tsao,Liu, & Kuhl, 1999), English-speaking adults performed significantly worse(77 % correct) than Mandarin-speaking adults (97 % correct) in an AXdiscrimination task when they were presented with a synthetic Mandarinalveolo-palatal fricative vs. affricate contrast. Thus, English-speakingadults experienced more difficulty in the discrimination of the non-nativeconsonant contrast. The present study employed the conditioned head-turnprocedure to examine the discrimination of English-learning andMandarin-learning infants at 6-8 and 10-12 months of age on the sameMandarin stimuli that had been presented to adults. We testedMandarin-learning infants in Taiwan and English-learning infants in the USwith the same experimental equipment, procedure and personnel. The primaryresults show that the older Mandarin-learning infants perform better thanthe younger Mandarin-learning. In contrast, the older English-learninginfants experience more difficulty in discriminating the non-nativecontrast than younger English-learning infants. The pattern of phoneticdiscrimination shown by infants from the two countries provides furtherevidence that experience listening to the native language alters theperceptual world of infants during the first year of life.
poster
Four experiments assess whether the perceptual properties offunction morphemes aid infants' speech segmentation abilities. In theseexperiments, the stress and phonological characteristics of functors aresystematically varied to determine their contribution to word segmentation.In a head-turn preference procedure, 11-month-old English-hearing infantsare first familiarized with two sentence-like utterances. One possessesthe English functor, 'the,' in sentence-initial position. The otherpossesses a nonsense 'functor' in that position. This nonsense syllable isinconsistent with the stress and/or phonology of English determiners.Following familiarization, infants are presented four test trials: (a) asyllable, now in isolation, that had immediately followed the Englishfunctor during familiarization (the EF syllable); (b) the syllable that hadfollowed the nonsense functor (the NF syllable); and (c) two differentnovel syllables. If infants are using function morphemes to assistsegmentation, then the EF syllable should be recognized as familiar at testand, consequently, should be less interesting than the novel syllables. Ifa nonsense functor fails to aid segmentation, the NF syllable should betreated as novel, so should be more interesting than the EF syllable andshould not differ from the two novel syllables. In Experiment 1, the nonsense functors differed from Englishdeterminers on the two dimensions of stress and phonologicalcharacteristics. As predicted, infants looked significantly longer fornovel syllables than for EF syllables (F (1,31) 10.72, p < .01); no suchdifferences were observed for NF syllables. Infants also attended longerto NF than to EF syllables (F (1,31) 4.76, p < .05). These results areconsistent with the prediction that infants can use function words toassist in the segmentation of subsequent words or syllables. The remaining three experiments were designed to determine thespecific properties of function words that are responsible for theirutility as segmentation cues. Experiment 2 assessed the possible role ofstress: The nonsense 'functor' was unstressed but inconsistent withEnglish functor phonology. Preliminary results indicate that infants areshowing the same pattern of preferences as was observed in Experiment 1.Thus, it is not simply the case that infants are extracting any syllablethat follows a stressed syllable. Experiments 3 and 4 assess the role ofphonology: In Experiment 3, the nonsense 'functors' are stressed, butconsistent with English phonology; in Experiment 4, the nonsense 'functors'are consistent both with English stress and phonology. Results indicating that infants are using the perceptual propertiesof function morphemes to identify word boundaries would provide support forthe prosodic bootstrapping hypothesis, which proposes that infants becomeaware of important linguistic units--such as words, phrases andclauses--through identifying perceptual characteristics that correspond tothese units in their native language. Specifically, it would address aknown problem with the hypothesis: phrases in the speech stream do not havea reliable set of perceptual correlates. If infants are indeed sensitiveto function morphemes, they would have a very useful tool for identifyingphrases in the speech stream because function morphemes are predictableindicators of phrase boundaries.
poster
Newborn infants come equipped with a sophisticated auditory system, whichallows them to perceive fine grained phonetic distinctions. Nevertheless,despite the newborn's relatively mature ability to discriminate phoneticsegments, it appears that it is the suprasegmental or prosodic aspects ofspeech which capture their attention. Newborns prefer their mother's voice tosomeone else's mother's voice, and also prefer their native language to otherlanguages. Evidence suggests that both these preferences are learnt in the womband are based upon speech rhythm and prosody. This information is conveyed inthe low frequency region of the speech spectrum, and is available to infantswhen the auditory system becomes functional from the beginning of the thirdtrimester of gestation. On the other hand, phonetic detail is not discriminablethrough the low-pass filter of the mother's womb. In summary, newborns havespeech preferences based on prosody, not phonetic segments, despite theirfine-grained abilities with the latter. The experiments here concern infants'relative attention to suprasegmental (prosodic) and segmental (phonetic)information in infants' preference for their native language. English languageenvironment infants of 4 BD, 6, 7BD, and 9 months were tested for their relativepreference for two lists, one of Thai and one of English words. The criticaldifferences between English and Thai form an are ideal basis on which todetermine infants' native language preferences: Thai is tonal, having fivelexical tones, while English is not; on two-syllable words English has apredominantly a strong-weak stress pattern, while Thai has a predominantlyweak-strong stress pattern; and there are phonetic segments which are peculiarto each language. The availability of word stress was manipulatedwithin-subjects, by testing infants on lists of one-syllable, and lists oftwo-syllable words. In Experiment 1, English language environment infants weretested for English vs Thai word list preference in normal speech. In Experiment2, English language environment infants were again tested but this time withEnglish vs Thai word lists which were low-pass filtered, a process whichrenders phonetic detail unintelligible. Analysis of variance showed thatacross all ages, native language preferences only result when two-syllablewords are presented in normal speech. Thus, it appears that both familiarstress patterns and familiar phonetic segments are required for native languagepreferences at these ages. These results are discussed with respect to infants'developing attention to phonetic detail in their native language, and the roleof prosody in speech perception development, and language acquisition.
poster
Recent research (Christophe & Morton, 1998; Nazzi, Bertoncini, &Mehler, 1998) suggests that language discrimination between birth and 2months, previously accounted for by the recognition of the nativelanguage, results from infants' sensitivity to the rhythmic properties oflanguages. This sensitivity allows young infants to discriminatelanguages between, but not within, rhythmic classes (stress-, syllable-,and mora-based). Given that this rhythmic sensitivity results in thenon-discrimination of the native language from other languages from itsrhythmic class, a series of experiments was conducted to determine theevolution of infants' ability to discriminate languages. Three possibledevelopmental hypotheses were explored. The maturation hypothesispredicts that the emergence of within-class discrimination will depend onthe acoustic distance between the languages presented, and will beindependent of their relation to the native language and its rhythmicclass. The rhythmic class acquisition hypothesis states that infantslearn the general rhythmic properties of their native class. It predictsthat all languages within the native language class, but only thosewithin that class, will become discriminable. Finally, the nativelanguage acquisition hypothesis states that infants learn the specificproperties of their native language, which will allow them to performwithin-class discriminations when, but only when, one of the languagespresented is the native language (or one of its dialects).
The present study used a new modification of the head-turn preferenceprocedure to investigate language discrimination by English-learningAmerican 5-month-olds. Given the modification of the procedure, weinitially checked whether the procedure is adapted to the study oflanguage discrimination at that age, using languages from differentrhythmic classes. Accordingly, Experiment 1 showed that 5-month-oldsdiscriminate between a dialect of their native language (stress-basedBritish English) and a language from a foreign rhythmic class (mora-basedJapanese), and Experiment 2 extended this result to languages from twoforeign rhythmic classes (syllable-based Italian vs. Japanese). Thevalidity of the procedure being established, we tested the infants onfour within-class pairs. Experiment 3 contrasted two languages from theforeign syllable-based class (Italian vs. Spanish). Experiment 4-6, usinglanguages from the infants' native stress-based class, differed accordingto whether the native language, a variant of that language, or onlyforeign languages were presented (British English vs. Dutch; British vs.American English; Dutch vs. German).
The results show that 5-month-olds can perform within-classdiscriminations when either the native language, or a variant of thatlanguage, is presented (British English vs. Dutch; British vs. AmericanEnglish). They also show that infants cannot perform within-classdiscriminations when both languages presented are equally dissimilar tothe native language, whether they belong (Dutch vs. German) or not(Italian and Japanese) to the native rhythmic class. This pattern ofresults supports the native language acquisition hypothesis, and suggeststhat the ability to discriminate languages from the same rhythmic classis, at 5 months, based on infants' developing knowledge of the soundorganization of their native language.
poster
Recent studies (Shi, Werker & Morgan 98, 99) showed that neonatesand 6-month-olds are able to classify words into lexical and grammaticalcategories, the most fundamental categories in all human languages. Wefamiliarized infants to a list of either lexical or grammatical wordsexcised from maternal speech, and then tested them on either a new list ofwords from the contrasting category (experimental condition) or a new listof words from the same category (control condition). Neonates were testedin the High-Amplitude Sucking procedure. Six-month-olds were tested usingvisual habituation. The neonates showed robust recovery, evincingcategorical discrimination. The 6-month-olds showed asymmetrical recoveryin their looking time-- recovery to lexical words after habituation togrammatical words but not vice versa. This asymmetry indicates a possiblelistening preference for lexical words, suggesting that language experiencemay have exerted an influence on older infants' listening strategy. In the current study we conducted two experiments to directly testif 6-month-olds prefer to listen to lexical words over grammatical words.English-learning infants (12 for each experiment) heard alternating listsof lexical and grammatical words in a Central Fixation Preferenceprocedure. Their listening preference was measured by their looking time toa central checkerboard while listening to each category of words. Resultsof the first experiment reveal that infants listened significantly longerto lexical words than to function words. In the first experiment we used randomly selected lexical andgrammatical word tokens from a mother's spontaneous speech to her infant.As a result, the type/token ratio of lexical words was higher than that ofgrammatical words; moreover, the lexical word list contained bothmonosyllabic and bisyllabic words whereas grammatical word list containedonly monosyllabic words. To best reflect the properties of natural speechinput, we decided to keep these differences in the first experiment. Thisraises the possibility, however, that our results might be accounted for bythese differences. We therefore conducted a second experiment in which boththe number of types and syllables were balanced for lexical and grammaticalword lists. This second experiment showed the same pattern of results asthe first -- infants listened significantly longer to lexical words than togrammatical words. The findings of these experiments provide direct evidence that6-month-old infants prefer to listen to lexical words over grammaticalwords. This early-appearing preference for the less frequently occurring,but acoustically and phonologically more salient lexical words (Shi, Morgan& Allopenna 98) may represent an important step toward languageacquisition. It focuses infants' attention on just those words which areneeded to break into an analysis of meaning.
poster
The effect of accentual pattern on the recognition of familiar wordswas examined in a set of experiments using the head-turn preferencetechnique. To establish a baseline, we tested English-learning infantsaged 11 months, in the absence of any special training or contextualsetting, on their relative attention to a list of words used frequentlyin infants' homes ('familiar words') as compared with a phoneticallyand phonotactically comparable list of unfamiliar words. Like the11-month old infants acquiring French in an earlier study (HallE9 &Boysson-Bardies, 1994), the 11-month-olds were found to attend longerto the familiar words (p << .05). In contrast, 9-month olds failed toshow significantly different listening times to the two lists. HallE9 & Boysson-Bardies (1996), reporting that French infantsrecognized familiar words despite a change in the initial consonant butfailed to show significantly longer looks to familiar words withchanged medial consonants, suggested that children's early lexicalrepresentations may be global rather than fully specified. As afollow-up to this suggestion, we explored further the nature ofchildren's representation of familiar words with the specific goal ofdetermining the balance between attention to accentual pattern vs.segmental structure. In order to establish the role of prosodyEnglish-learning children were presented with a list of familiar wordswith altered stress (trochaic words given iambic stress and iambicwords or disyllabic phrases given trochaic stress) contrasted withunaltered familiar words. No significant difference was found,suggesting that the stress pattern was not an essential part of theinfants' early lexical representations. We validated this finding byconducting an additional experiment in which we contrasted the samelist of altered familiar words used in the previous experiment withaltered unfamiliar words; under these conditions the former receivedsignificantly longer looks (p <<.001). The last pair of experiments was designed to test whether accentedsyllables are more fully specified in infant word representations thanunaccented syllables. We predicted that there would be a differencebetween infants' representation of early French words, which areiambic, and early English words, which are largely trochaic. Wepresented English-learning infants with lists of familiar andunfamiliar (trochaic) words with a change in manner of articulation of,first, the medial consonants and, in a second experiment, the initialconsonants. As anticipated, the infants attended longer to the familiarwords despite the change in the second consonant (p << .01).Preliminary findings from the second experiment suggest that attentionto the familiar words with a changed first consonant will not reachsignificance. We conclude, first, that at 11 months segmental pattern plays a moreimportant role in infants' mental representation of words than prosodicpattern. Secondly, the accentual pattern of the adult language mayinfluence the level of detail in infant early word representations. Indisyllabic words infants acquiring both English and French appear tounderspecify the onset consonant of an unstressed syllable, theword-initial consonant in the case of French, the medial consonant inthe case of English.
poster
Decades of research have shown than infants have remarkably acutespeech perception skills early on in life, skills that enable them to makefine phonetic discriminations in both native and unfamiliar languages.Whereas these findings initially bolstered arguments for cerebralspecialisation for language, recent studies have revealed that many ofthese discrimination abilities are not exclusive to phonetic perception,but are also harnessed by auditory processors to discriminate betweennon-speech sounds. For example, the difference in voice onset time thatmakes /da/ acoustically discernible from /ta/ to an infant listener, alsoallows the infant to discriminate between pairs of tones with equivalentdifferences in onset time. Given that auditory perception abilities are somewhat promiscuous,are there initial processing biases that facilitate infantsED attention tospeech, which might, in turn, facilitate the acquisition of additionalspeech properties? In this series of experiments, we investigated infantsEDpreference for listening to speech over matched non-speech sounds, as onepotential initial processing bias.One difficulty in addressing this question lies in the nature of the speechsignal itself. While previous studies have compared preference for speechto white noise or music, the non-speech stimuli we used are designed to bemoderately faithful to their speech counterparts. Speech tokens consistedof nonsense words with a CVC (consonant-vowel-consonant) structure.Non-speech sounds were derived from nonsense words by extracting andreplacing the fundamental frequency and the first three broad-band formantsof our speech tokens by simple sinusoidal waves tracking each waveEDs centrefrequency. We tested infants of two ages, 4-5 and 6-7 months, for preferencefor speech over non-speech using a sequential preferential lookingprocedure. In this infant-controlled procedure, alternating strings ofspeech and non-speech were paired with a visual display, a black and whitecheckerboard. Infant fixation to the display for each type of sound wassummed and compared for preference. Results indicate that infants show aclear preference for listening to speech over sine-wave counterparts by 6-7months of age. Preference for speech over non-speech appears to beemerging in 4-5 month-old infants. These results suggest that infants of6-7 months are sensitive enough to the microstructure of speech toselectively attend to speech over highly similar non-speech sounds. Thisselective orientation facilitates infantsED attention to the internalfeatures of speech. Is the emerging preference seen in the younger infants due to alack of discrimination between the speech and non-speech sounds? Or do thecharacteristics shared by both types of signals satisfy the youngerinfantEDs requirements for a speech-like sound? Do older infants engagemore stringent criteria for what speech-like stimuli they will selectivelyattend to? We are currently addressing these questions by investigatinginfantsED discrimination abilities and their attention to other speech-likestimuli.
poster
Learning accounts of language acquisition rely heavily on frequent cues inlinguistic input. Grammatical morphemes like 'the' and 'a' are frequentlypositioned before nouns, whereas '-ed' is often affixed to verbs.Frequently occurring cues also mark longer distance dependencies, such asagreement between auxiliaries and 'ing' endings as in 'the girl isrunning.' Such regularities are thought to act as anchors in the input,signaling syntactic phrases. But, do such regularities really aidlearning? If so, under what conditions? One possibility is that the morefrequently cues occur, especially in more variable contexts, the morelikely they are to be perceived as invariants. If so, environments withmore variability should promote greater learning.Infants (age 12 and 17 months) were exposed to auditory strings produced byone of two grammars. One grammar produced sentences of the form: S --> AXC and S --> BXD,the other produced sentences: S --> AXD and S --> BXC.The middle element (X) was drawn from a set of 12 or 18 words. A-D werefixed, one-syllable words. Xs were 2-syllable words. For instance, legalstrings in one grammar were PEL-WADIM-RUD and VOT-WADIM-JIC and in theother were PEL-WADIM-JIC and VOT-WADIM-RUD. Because X elements occurred inboth grammars, learners could only distinguish the grammars by learning thedependencies between the first and third elements. The set size of X wasmanipulated to investigate the effect of variability of context on learning(in the low variability condition each instance of X was drawn from one of12 items, in the high variability condition it was drawn from one of 18).We tested 12- and 17-month-olds because of prior work showing sensitivityof older, but not younger, infants to dependencies separated by oneintervening item (Santelmann & Jusczyk, 1998). Such dependencies areparticularly difficult for learners to acquire.Infants were exposed to strings from their training grammar for threeminutes using the head-turn preference procedure. They then listened tostrings from their training grammar and strings from the other grammarduring test. Infants who had learned the dependency between the first andthird element should exhibit different mean looking times to stringsviolating the dependency. The dependencies between the first and thirdelements were identical in the low- and high- variability conditions. Theonly difference was in the size of the set from which irrelevantintervening elements were selected. Infants relying on rote memory foradjacent elements should perform worse in the high- than in the low-variability condition. Those able to capitalize on invariant cues shouldperform better in the high-variability condition.Seventeen-month-olds distinguished legal from illegal dependencies in thehigh- but not in the low-variability condition (12-month-olds failed todistinguish dependencies in either condition). The result that17-month-olds perform better under conditions of higher memory load iscounterintuitive, demonstrating that infants this age can forgo processingof adjacent elements as other sources of information become more salient(as when frequently occurring cues are perceived as invariant because ofembedding in highly variable contexts).
poster
Even in infant-directed speech, parents expose their children to prosody that serves a variety of communicative functions, such as focusing stress that indicates the given/new distinction (Fisher & Tokura, 1995). Although it has been demonstrated that children are able to segment syllabic strings based on their rhythmic structure, this has been done using highly controlled, artificial stimuli that carry few of the many supralexical features characteristic of fluent conversational speech. In a series of studies, we examine how the natural prosodic marking of the given/new distinction influences infants' use of speech rhythm in identifying multisyllabic words.We began with acoustic analyses of the given/new stress present in naturally produced infant-directed speech. Our procedure for collecting data in this initial study also provided speech stimuli for use in subsequent perceptual studies. We used a cued puppet show (Fisher and Tokura, 1995) to elicit new- and given-stressed tokens of specific words. Mothers watched a puppet show with their infants and were instructed to describe the events enacted in the puppet show to their infants as they occurred. Puppet show descriptions were elicited from eight mothers of 9-month-olds. Adult listeners rated sentences from the mothers' descriptions on a 7-point Likert scale for how given or new the stress on the target words within the sentences sounded. Those sentences containing target words that were unambiguously rated one way or the other were then mimicked by an actor (to control for speaker effects) for use in a series of perceptual studies.For the perceptual studies, two groups of 24 English-exposed infants (7 and 10 months of age) were tested on their ability to detect familiarized disyllabic words in fluent speech. Stimuli consisted of English target words (for familiarization sessions) and sentences (for testing sessions), selected in the manner outlined above. A key manipulation between experiments was the age of infants. Within each experiment, familiarization items carried either all given or all new stress, or alternated between the two. Mean listening times were higher for new-stressed stimuli than for given-stressed stimuli across both ages, reflecting the superior attention-attracting and -maintaining qualities of new-stressed words. Ten-month-olds listened longer to given-stressed stimuli than did 7-month-olds. However, 7-month-olds listened longer to stimuli that alternated between given and new stress than they did to given-stressed stimuli alone. Our earlier acoustic analyses indicated that mothers alternate naturally and consistently between new and given stress when speaking to their infants; we discuss this finding in terms of distributed learning. Overall, these results validate the notion that infants must hear a word several times before they can recognize that word in subsequent usage (Jusczyk & Aslin, 1995), further highlighting the importance of repetition in infant-directed speech.