Near-term fetuses process temporal features of speech

Developmental Science 14:2 (2011), pp 336–352

DOI: 10.1111/j.1467-7687.2010.00978.x

PAPER Near-term fetuses process temporal features of speech Carolyn Granier-Deferre,1 Aure´lie Ribeiro,1 Anne-Yvonne Jacquet2 and Sophie Bassereau1 1. Laboratoire Psychologie et Neurosciences Cognitives (CNRS UMR 8189), Universit Paris Descartes, Institut de Psychologie, France 2. Laboratoire Psychologie de la Perception (CNRS UMR 8158), Universit Paris Descartes, France

Abstract The perception of speech and music requires processing of variations in spectra and amplitude over different time intervals. Nearterm fetuses can discriminate acoustic features, such as frequencies and spectra, but whether they can process complex auditory streams, such as speech sequences and more specifically their temporal variations, fast or relatively slow acoustic variations, is unclear. We recorded the cardiac activity of 82 near-term fetuses (38 weeks GA) in quiet sleep during a silent control condition and four 15 s streams presented at 90 dB SPL Leq: two piano melodies with opposite contours, a natural Icelandic sentence and a chimera of the sentence – all its spectral information was replaced with broadband noise, leaving its specific temporal variations in amplitude intact without any phonological information. All stimuli elicited a heart rate deceleration. The response patterns to the melodies were the same and differed significantly from those observed with the Icelandic sentence and its chimera, which did not differ. The melodies elicited a monophasic heart rate deceleration, indicating a stimulus orienting reflex while the Icelandic and its chimera evoked a sustained lower magnitude response, indicating a sustained attentional response or more focused information processing. A conservative interpretation of the data is that near-term fetuses can perceive sound streams and the rapid temporal variations in amplitude that are specific to speech sounds with no spectral variations at all.

Introduction The perception of speech and music melodies requires processing of variations in spectra and amplitude over different intervals of time (temporal windows). Pitch variations over time and specific pitch interval ratio relationships, thus spectral variations, are critical features of music. This is less true of speech. In music, these variations are smaller than in intonation contours in speech. They also occur over much larger time scales, of the order of hundreds of ms, than the phonetic categories in speech. In contrast, very rapid temporal variations, of the order of tens of ms, are ubiquitous in speech and virtually absent from music (Zatorre, Belin & Penhune, 2002). Near-term fetuses can discriminate differences in sound pressure levels, frequencies and spectra, which are necessary for perceiving melodic contours and prosodic features in speech. However, whether they can process auditory streams, and more specifically the temporal envelope, i.e. the relatively slow variations in amplitude over long intervals (‡ 150–300 ms), and the variations over brief intervals (< 40 ms), i.e. the temporal fine structure of speech, in utero, is unclear. Such temporal

processing is necessary for the discrimination of speech features (Shannon, Zeng, Kamath, Wygonski & Ekelid, 1995; Smith, Delgutte & Oxenham, 2002; Zatorre et al., 2002). In the adult, envelope is considered critical for the perception of contours and prosodic cues, while the temporal fine structure is critical for phonetic discriminations, in particular for perceiving stop consonants (Mirman, Holt & McClelland, 2004; Poeppel, 2003). Children also use temporal cues in speech discrimination (e.g. Bertoncini, Serniclaes & Lorenzi, 2009; Lorenzi, Dumont & Fllgrabe, 2000). Several fetal and newborn studies on the effects of prenatal familiarization with speech and music indicate that the fetus can perceive auditory streams and, thus, that temporal processing should occur in utero. Recently, it was shown that neonatal learning is constrained by the temporal features of the auditory reinforcers (DeCasper & Prescott, 2009). We review this literature below. However, the development of temporal processing itself has never been examined. The present research investigated the near-term fetus’ perception of streams of music melodies and streams of speech, which requires the integration of acoustic information over different temporal windows.

Address for correspondence: Carolyn Granier-Deferre, Laboratoire Psychologie et Neurosciences Cognitives (CNRS UMR 8189), Universit Paris Descartes, Institut de Psychologie, 71 avenue Edouard Vaillant, Boulogne-Billancourt, 92 – France; e-mail: [email protected]

 2010 Blackwell Publishing Ltd, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA.

Fetuses process temporal features 337

Auditory development Early anatomical and neuro-functional development of the auditory apparatus has been extensively studied in animal models. The data have been used to estimate that the onset of cochlear function in humans occurs at around 24 weeks gestational age (GA) (Pujol, Laville-Rebillard & Uziel, 1991; see reviews in Pujol, Laville-Rebillard & Lenoir, 1998; Pujol & Uziel, 1988; Rbsamen & Lippe, 1998). Data obtained from behavioral studies in full-term infants (for review see Werner & Gray, 1998), together with the electrophysiological data at all levels of the auditory system in the premature infant, provide a clear picture of human auditory development. The main components of brainstem auditory evoked potentials and some components of cortical auditory evoked potentials can be reliably recorded at 25–26 weeks GA. This is followed by an increase in the number of components that are recorded, a regular increase in the amplitude of the components, a decrease in the latencies between components, and a progressive decrease in absolute auditory thresholds. By 35 weeks GA, cochlear biomechanics and frequency selectivity are mature (Morlet, Collet, Salle & Morgon, 1993; Morlet, Lapillone, Ferber, Duclaux, Sann, Putet, Salle & Collet, 1995). One can consider that all features of the acoustic signal are transmitted along the auditory pathways to the auditory cortex, with similar thresholds to the term newborn but longer latencies (e.g. Starr, Amlie, Martin & Sanders, 1977; Krumholz, Felix, Goldstein & McKenzie, 1985; Pasman, Ntanen & Alho, 1991; Rotteveel, Stegeman & de Graaf, 1987b; Rotteveel, de Graaf, Stegeman, Colon & Visco, 1987a; Eldredge & Salamy, 1996; Moore, Ponton, Eggermont, Wu & Huang, 1996). That human fetuses respond reliably to loud airborne noises > 100 dB SPL (re: 20 lPa) with startle movements and cardiac accelerations from 28 weeks GA onward has been well documented, beginning with the pioneering studies in the early 20th century (Forbes & Forbes, 1927; Peiper, 1925; Ray, 1932; Sontag & Wallace, 1935, 1936; Spelt, 1938; see review of the first studies in Busnel & Granier-Deferre, 1983; Kisilevsky & Low, 1998). Many studies were carried out for clinical purposes and used vibro-acoustic stimuli (Birnholz & Benacerraf, 1983; see review in Kisilevsky, 1995; Kisilevsky, Muir & Low, 1992). However, the latter, because they also provide somesthetic and proprioceptive stimulations, do not give information on audition per se. Fewer fetal studies used airborne auditory stimuli or controlled the presence of low frequency vibrations. In the near-term fetus (‡ 36 weeks GA), reactions to airborne sounds are quite similar to the ones found in the newborn. They vary as a function of sound pressure level (SPL), frequency and bandwidth (see reviews in Busnel, Granier-Deferre & Lecanuet, 1992; Lecanuet, Granier-Deferre & Busnel, 1995), and with behavioral states (Nijhuis, Prechtl, Martin & Bots, 1982), as well (Groome, Mooney, Holland, Smith, Atterbury & Dykman, 1999; Granier-Def 2010 Blackwell Publishing Ltd.

erre, Lecanuet, Cohen & Busnel, 1985; Lecanuet, Granier-Deferre, Cohen, Le Houezec & Busnel, 1986). At lower SPL, £ 100 dB, the cardiac response to airborne sounds changes from acceleration to a low amplitude deceleration without accompanying movements (Groome et al., 1999; Lecanuet, Granier-Deferre & Busnel, 1988, 1989, 1995). This same shift occurs in newborns at similar relative SPL, i.e. with about 25 dB difference which accounts for the acoustical isolation of the fetus from external sounds. In newborns and infants, heart rate (HR) decelerations are interpreted as orienting responses, reflecting information processing, and perception. They are considered correlates of orientation or attention toward the stimulus (Berg & Berg, 1987; Clarkson & Berg, 1983; Graham, 1992; Graham, Anthony & Zeigler, 1983; Richards, 1997). Since about 2000, functional magnetic resonance imaging (fMRI) (e.g. Moore, Vadeyar, Fulford, Tyler, Gribben, Baker, James & Gowland, 2001) and magnetoencephalography (fMEG) (e.g. Eswaran, Lowery, Robinson, Wilson, Cheyne & McKenzie, 2000; Holst, Eswaran, Lowery, Murphy, Norton & Preissl, 2005; Porcaro, Zappasodi, Barbati, Salustri, Pizzella, Rossini & Tecchio, 2006; Schleussner & Schneider, 2004) have been used to examine prenatal auditory function. General results are in agreement with the previous fetal studies, but like the heart rate data, tell us little about what the fetus can actually process and perceive about auditory streams inside the amniotic fluid. Most information about the near-term fetus’ perception of complex sounds and auditory streams is indirect, coming from learning studies in the newborn. More than 20 years ago it was experimentally demonstrated, with operant non-nutritive sucking choice procedures (see DeCasper & Spence, 1991), that 2–3-day-old newborns recognize and prefer speech and melodic sequences that their mothers spoke (DeCasper & Spence, 1986) or sang (Cooper & Aslin, 1989) repeatedly while pregnant. Significantly, it made no difference whether the mother or another woman read the story or sang the melody during the postnatal test. Furthermore, newborns prefer their mother’s language, either English or Spanish, to the one they have never heard before (Moon, Cooper & Fifer, 1993). Later, Mastropieri and Turkewitz (1999) showed that neonates can discriminate the same sentence pronounced with an angry voice or not, but only if the sentence is in the maternal language. Learning and preference for jazz or classical melodies that the mothers listened to while pregnant was also shown (Woodward & Guidozzi, 1992). Hepper (1988, 1991) demonstrated that such learning can occur naturally in daily life. Newborns went quickly into a quiet alert attentive state upon hearing the jingle of a soap opera that their mothers watched daily during their pregnancy, and not upon hearing the backward version of the jingle. The most conservative explanation of these data is that newborns possess recognition memory of the spectral and temporal features of recurrent prenatal stimulation. Thus, the term

338 Carolyn Granier-Deferre et al.

fetus has the capacity to process some spectral variations over time, i.e. prosody, melodic contour of speech, language, and music, and possibly, some phonetic features of the mother’s language, as well. Considering the precocity, the progressive nature and the plasticity of the functional development of the auditory system, the perceptual abilities demonstrated in newborns are unlikely to have blossomed at birth (Granier-Deferre & Lecanuet, 1987; D. Moore, 2002; J. Moore, 2002). If so, prenatal auditory experience makes an important contribution to the development of the newborn’s capacities for speech perception (Lecanuet & Granier-Deferre, 1993; Lecanuet, Granier-Deferre & Busnel, 1991; Fifer & Moon, 1994, 1995; Moon et al., 1993; Moon & Fifer, 2000). However, most theories about the early stages of speech ⁄ language perception in infants and its subsequent development fail to consider, or minimize, the potential contribution of prenatal hearing and perception to the development of speech perception after birth (see for example, Dehaene-Lambertz, Hertz-Pannier & Dubois, 2006; Vouloumanos & Werker, 2007, and commentaries in Rosen & Iverson, 2007). The first phase of the native language magnet model developed by Kuhl (1991, 2000; Kuhl, Conboy, Coffey-Corina, Padden, Rivera-Gaxiola & Nelson, 2008) does acknowledge the contribution of prenatal experience, but the results of the neonatal research cited above imply a greater contribution than is proposed. Perhaps prevalent theories simply underestimate the amount, the nature, and the vast array of internal and external sounds, including speech, that activate the fetal auditory system for several months before birth. Characteristics of sounds available to the fetal ear Numerous studies over the past 50 years have assessed the maternal intrauterine sound environment (see review of the first studies in Busnel & Granier-Deferre, 1983; Lecanuet et al., 1995). Since the 1980s, power spectrum analyses of intra-amniotic recordings in women during delivery (Benzaquen, Gagnon, Hunse & Foreman, 1990; Gagnon, Benzaquen & Hunse, 1992; Querleu, Renard, Boutteville & Crpin, 1989; Querleu, Renard & Crpin, 1981; Querleu, Renard, Versyp, Paris-Delrue & Crpin, 1988a; Querleu, Renard, Versyp, Paris-Delrue & Vervoot, 1988b) and in the gestating ewe (Abrams, Gerhardt & Peters, 1995; Armitage, Baldwin & Vince, 1980; Gerhardt, 1989; Peters, Abrams, Gerhardt & Griffiths, 1993; Richards, Frentzen, Gerhardt, McCann & Abrams, 1992; Vince, Billing, Baldwin, Toner & Weller, 1985) have provided a relatively clear picture of the characteristics of the sounds that can stimulate the maturing fetal auditory system. It should be mentioned here that cardio-vascular noises are louder during delivery than during gestation (Armitage et al., 1980; Vince et al., 1985). Thus, human recordings that were made during delivery underestimate the signal ⁄ noise ratio of the fetus’ daily sound environment. Overall, these studies show that the fetal ear can be  2010 Blackwell Publishing Ltd.

stimulated by all maternal biological sounds, respiratory, gastro-intestinal, cardio-vascular, laryngeal, and noises from movements. The pulsed maternal and placental vascular noises are constantly present, but other sounds like maternal vocalizations or digestive gurgles occur recurrently. Thus, different auditory streams, with their own spectra and envelope, are present simultaneously or vary in time of occurrence. SPL is not homogeneous inside the amniotic fluid. Human recordings during delivery (Querleu et al., 1981, 1988a, 1988b, 1989; Benzaquen et al., 1990) show that the background noise is louder closer to the placenta. SPL decreases as frequency increases, from about 30–60 dB in its lowest frequencies, < 100 Hz, to only 10 dB in its highest frequencies, > 500– 700 Hz. In sum, the mean SPL of the maternal background noise is comparable to that found in a quiet room and it does not mask external sounds above 40–50 dB. Because of mismatched impedances, mainly between air and biological media, some components of external sounds are attenuated inside the amniotic fluid. SPL attenuation that has been measured both in humans (Querleu et al., 1981, 1988a, 1989; Richards et al., 1992) and in sheep (Abrams et al., 1995; Armitage et al., 1980; Gerhardt, 1989; Lecanuet, Gautheron, Locatelli, Schaal, Jacquet & Busnel, 1998; Peters et al., 1993; Vince et al., 1985) is affected by the quality of the transducer and its location inside the amniotic fluid. In general, attenuation of pure tones and band noises depends on frequency: Frequencies up to about 500 Hz are not attenuated. Attenuation increases with frequency at a rate of about 6 dB ⁄ octave and does not exceed 30 dB at 4 kHz. In pregnant women, Querleu et al. (1981, 1988ab, 1989) reported no more than 28 dB attenuation at 2 kHz and Richards et al. (1992) describe even lower attenuation values and large individual differences (15–20 dB). Higher frequencies undergo highly variable non-linear attenuation when passing from one medium to another (Oliver, 1989). This has been reported in the ewe, where high frequency attenuation can be less than medium frequency attenuation (Lecanuet et al., 1998; Peters et al., 1993; Vince et al., 1985) and the presence of standing waves can produce SPL reaching external levels or higher (Lecanuet et al., 1998). Data from the fetal guinea pig show that aerial free-field pure-tone stimulations up to 20 kHz induce a marked increase in (14-C) 2-deoxyglucose uptake in the cochlear nucleus and in the inferior colliculus. Frequency-specific auditory labeling was found for all frequencies of human hearing (Horner, Servire & Granier-Deferre, 1987; Servire, Horner & Granier-Deferre, 1986). The transfer functions of complex sounds are more complicated, but their overall attenuation is lower than that of tones and band noises. For example, music attenuation is £ 10 dB SPL in the gestating ewe (Abrams, Griffiths, Huang, Sain, Langford & Gerhardt, 1998; Gerhardt, 1989). In humans, intra-amniotic recordings of external voices emitted at 60–70 dB SPL show an overall attenuation of 20 dB, with no significant difference

Fetuses process temporal features 339

between male and female voices (Benzaquen et al. 1990; Querleu et al., 1988b, 1989). Voicing, including laughing, emerges clearly from the background noise. Individual voices, prosody and many phonetic features are preserved and recognizable by adults. Spectrograms show that F0 and the first formants of vowels are well transmitted (Querleu et al., 1988b, 1989). Intelligibility of external speech has been examined with audiometric lists of meaningful and non-meaningful speech stimuli. Some phonetic categories are better recognized than others. Significantly, consonants are better recognized in maternal speech. In addition, vowels sharing the same first formant were discriminated the least (Querleu et al., 1988b, 1989). Overall, 30% of the phonemes are recognizable, whatever the voice (maternal or external). Similar recognition rates were found for recordings made in the gestating ewe. Recognition is significantly affected by presentation levels (Griffiths, Brown, Gerhardt, Abrams & Morris, 1994), and by the recording site of the transducer (see below). The maternal voice is transmitted internally to the amniotic fluid via body tissues and bones and partly through some of its aerial components. F0 and the first overtones are fully conducted through the spine and the pelvic arch (Petitjean, 1989). Comparison of SPL of the maternal voice measured in air and inside the amniotic fluid shows little difference. Some investigators report a small attenuation, £ 8 dB (Benzaquen et al., 1990; Querleu et al., 1988b, 1989), and others a small amplification (Richards et al., 1992). It appears from this corpus of data that all frequencies within the human auditory range can be transmitted into the amniotic fluid with no more than 25–30 dB SPL attenuation and that many acoustic features of speech, emitted at and above conversational level near a pregnant mother, emerge from the background noise. Therefore, the developing auditory system can be activated by normal maternal speech sounds, and by the speech of others if spoken loudly enough. Maternal vocalizations alone expose the fetal auditory system to an array of speech variations in spectra and amplitude similar to those the newborn will later encounter out of the womb. Interestingly, a simple neural network can acquire the capacity to distinguish English stop consonants spoken by anyone, after mere exposure to exemplars from one speaker that were altered to have the characteristics of speech in utero (Seebach, Intrator, Lieberman & Cooper, 1994). The question of whether the mechanism of sound transmission from the amniotic fluid to the fetal cochlea, i.e. conduction via tissues and bone alone and ⁄ or via the tympano-ossicle chain, might affect fetal perception is often raised. A series of studies addressed this issue with technically difficult experiments in the fetal lamb. Some studies were designed to assess the complete ‘acoustic isolation’ of the fetus from external sounds, i.e. the SPL increase necessary to get equal amplitude cochlear microphonics (CM) inside and outside the uterus with  2010 Blackwell Publishing Ltd.

tone bursts delivered in air. Acoustic isolation measured through the CM was found to be greater than the amount of attenuation of the tone bursts in the uterus, for all frequencies tested (Gerhardt, Otto, Abrams, Colle, Burchfield & Peters, 1992; Gerhardt, Huang, Arrington, Meixner, Abrams & Antonelli, 1996). The general conclusion was that the fetus hears through bone conduction, and thus that the middle ear was non-functional in utero. Other authors reached the same conclusion (Sohmer, Perez, Sichel, Priner & Freeman, 2001). The characteristics of speech sounds from the CM recorded in utero and ex utero were also examined (Huang, Gerhardt, Griffiths & Abrams, 2002; Smith, Gerhardt, Griffiths, Huang & Abrams, 2003). For example, with speech sentences presented externally at 95 dB SPL, Smith et al. (2003) found different intelligibility rates according to the site of the recordings: almost 99% from inside the uterus, 73% from the CM of externalized fetuses (middle ears were cleared of all liquid and mucosa), and 41% from the CM after the middle ear was refilled with amniotic fluid and the fetuses returned to the womb. Unfortunately, the authors note several difficulties linked to the procedures which leaves the issue still unresolved. However, there are two reasons why the ultimate resolution of this issue will have no substantial impact on questions about fetal auditory perception. The first is empirical: All maternal sounds and external low frequencies (corresponding to the first four octaves of the piano) are not attenuated in utero and bone conduction, as found in conductive hearing loss, does not impair speech intelligibility. Thus, from the time the fetal cochlea is almost fully mature, about 32 weeks GA, it will be activated by a large array of sounds ‡ 50 dB SPL, including most speech features. The second is theoretical (Granier-Deferre, Lecanuet, Cohen, Busnel, Querleu & Sureau, 1984): In air, the tympano-ossicle chain gives the amplification needed (30 dB SPL) to compensate for the impedance mismatch between air in the ear canal and middle ear and the fluids inside the cochlea. In utero, there is no air; hence, there is no need for a functional middle ear. Because liquids, tissues, and bones have almost the same impedance, SPL inside the cochlea should be relatively similar to SPL inside the amniotic fluid. Fetal auditory perception Several studies have demonstrated that near-term fetuses are capable of making intensity, frequency and spectral based discriminations. Various cardiac and motor habituation ⁄ dishabituation procedures show that they can discriminate two pure tones, 250 Hz and 500 Hz (Shahidullah & Hepper, 1994), and two low-pitched piano notes, D4 and C5 (Lecanuet, Granier-Deferre, Jacquet & DeCasper, 2000). They also discriminate a male and a female voice contrasted on fundamental frequency, spectra, and timbre (Lecanuet, Granier-Deferre, Jacquet, Capponi & Ledru, 1993). Other studies have focused

340 Carolyn Granier-Deferre et al.

specifically on discrimination of vowels, which indicate that near-term fetuses can process spectra. They can discriminate the French vowels ⁄ a ⁄ and ⁄ i ⁄ , which have no common formants, as they appear in [ba] [bi] versus [bi] [ba] (Lecanuet, Granier-Deferre, DeCasper, Maugeais, Andrieu & Busnel, 1987), [baba] versus [bibi] (Shahidullah & Hepper, 1994) and ⁄ æ ⁄ versus ⁄ i ⁄ (Groome, Mooney, Holland, Bentz & Atterbury, 1997). Note that fetal reactivity to these vowel sounds was found as early as 26 weeks GA (Zimmer, Fifer, Kim, Rey, Chao & Myers, 1993). The fact that fetuses can detect frequency changes has also been shown with fetal MEG recordings of the Mismatch Negativity response (MMN) (Draganova, Eswaran, Murphy, Huotilainen, Lowery & Preissl, 2005; Draganova, Eswaran, Murphy, Lowery & Preissl, 2007; Huotilainen, Kujala, Hotakainen, Parkkonen, Taulu, Simola, Nenonen, Karjalainen & Ntnen, 2005). This cortical response reveals the detection of an odd auditory stimulus appearing within a series of the same standard sound. It is present 2–3 weeks after the onset of cochlear function and before cochlear biomechanics is mature. The MMN response to tone bursts, 500 Hz vs. 750 Hz, was observed in fetuses as young as 28 weeks GA (Draganova et al., 2007). MMN latency rapidly decreases with gestational age: 288 ms at 26–29 weeks GA and 197 ms at 34–37 weeks GA (Holst et al., 2005). Auditorially evoked cortical activation has been confirmed with fMRI studies. One study revealed left primary auditory cortex activation to modulated tones as early as 33 weeks GA (Jardri, Pins, HoufflinDebarge, Chaffiotte, Rocourt, Pruvo, Steinling, Delion & Thomas, 2008). Other studies using a vibro-acoustic stimulus found localized activation in the temporal and frontal lobes in the near-term fetus (Fulford, Vadeyar, Dodampahala, Ong, Moore, Baker, James & Gowland, 2004; Moore et al., 2001). The perception of differences in pitch or spectra can occur within the first hundreds of milliseconds after stimulus onset. To our knowledge, there is no clear information on fetal processing of either sound duration (see review in Lecanuet et al., 1995) or temporal variations over time. However, we know from electrophysiological studies in premature neonates that the auditory system responds to acoustic transients (clicks), i.e. very fast temporal changes like those that are important for phonetic discrimination, as early as 26 weeks GA. The repetition rate that the auditory system can integrate increases with maturation, especially after 30–32 weeks GA (for example, see Moore et al., 1996; Rotteveel et al., 1987a, 1987b). Animal studies show that clicks delivered in air are processed in the womb. Cochlear, brainstem, and cortical auditory evoked potentials recorded in the chronically implanted fetal guinea-pig (Scibetta & Rosen, 1969) and fetal sheep (Pierson, Gerhardt, Griffiths & Abrams, 1995; Woods & Plessinger, 1989) in utero have the same characteristics and same developmental course as the ones recorded in prematurely delivered animals.  2010 Blackwell Publishing Ltd.

Only a few studies have examined the discrimination of temporal variations in sound streams by the near-term fetus. They used cardiac habituation ⁄ dishabituation procedures. Manipulations of continuous and pulsed pure tones and of the phonemes ⁄ æ ⁄ and ⁄ i ⁄ show that changing the temporal characteristics of the stimulation is more effective in eliciting a new cardiac deceleration than changing its spectral complexity (Groome et al., 1999; Groome et al., 2000). Lecanuet, Jacquet and Bontemps (2002) reported discrimination of several temporal changes, for example, a 10% increase or decrease in the 600 ms inter-onset interval of an isochronous sequence of a musical note. Kisilevsky, Hains, Jacquet, Granier-Deferre and Lecanuet (2004) showed that the cardiac change during a standard lullaby differed from the one elicited with an accelerated version. Because the amount of energy activating the cochlea is not the same when a stimulus is pulsed, accelerated or decelerated, one cannot be certain that the modification in cardiac responsiveness is caused by the change in the temporal properties of the stimulus. The fact that fetuses can process variations in spectra and amplitude over time can be inferred from studies that have examined the effect of prenatal repeated exposure to speech streams on cardiac activity. Kisilevsky and colleagues found that the fetal cardiac response patterns to the maternal voice and to a female stranger’s voice were different (Kisilevsky, Hains, Lee, Xie, Huang, Ye, Zhang & Wang, 2003; Smith, Dmochowski, Muir & Kisilevsky, 2007). A fetal learning experiment found that the near-term fetus’ HR decreased to a recording of a short story that their mothers recited aloud for 4 weeks but not to a control story (DeCasper, Lecanuet, Busnel, Granier-Deferre & Maugeais, 1994). The two stories were emitted at a low level that does not usually produce an HR change and the speaker’s voice was irrelevant during the test. This result was later replicated in 32-week-old fetuses (Krueger, Holditch-Davis, Quint & DeCasper, 2004). These experiments show directly what the neonatal research implied, i.e. that the fetal auditory system can not only process spectral information, but that the fetus can recognize patterns of temporal variations in a speech stream, independently of the speaker’s voice properties. The preceding review shows the early functional development of the auditory system, that all sound features, including the spectral and temporal properties of speech, are present in utero, and that near-term fetuses can perceive spectral information. Information about their perception of temporal characteristics is less extensive. The present research focused specifically on fetal processing of temporal variations in auditory streams. We investigated whether the near-term fetus can perceive the envelope, i.e. the slow variations in amplitude involved in the processing of melodic contours and prosody, and ⁄ or the fast oscillations, i.e. the fine temporal structure of the stream, which is required for accurate speech perception. We used two categories of

Fetuses process temporal features 341

auditory streams, two different piano melodies, one with an ascending contour, the other with a descending contour, and two different speech sequences. One speech sequence was an Icelandic language sentence. Icelandic belongs to a different rhythmic class from the one spoken by the mothers (French, a syllabic language). It is a stress language, and it is rare, which ensured that its effect on fetal responsiveness cannot be confounded with familiarity. The second speech stream was a chimera of the Icelandic sentence. It was made by replacing its spectral content with white noise and retaining the same fast and slow amplitude variations that are present in the original sentence. Speech sequences and music melodies differ categorically in the nature of their temporal variations over time: The speech stimuli contain fast temporal amplitude changes, £ 40 ms, and the music melodies have none. All fetal subjects were 38 weeks GA, therefore had a fairly mature auditory system, and were tested during behavioral state 1F (Nijhuis et al., 1982). Each auditory stream was presented for 15 s to separate groups of fetuses. Our analysis focused on the latency, direction, magnitude, and duration of HR changes that occurred before and throughout the stimulation. Assessing the HR changes within and between categories can reveal the effects of temporal processing per se. If the fetus processes mainly frequency and spectra variations over time, therefore slow amplitude variations, then the heart rate change should differ between the chimera and the three other streams. If the presence versus absence of fast temporal amplitude variations significantly affects processing, then the characteristics of the cardiac change should differ between the music melodies and the speech streams and not between the two speech streams. Such a result would indicate that fetuses could process the fast temporal variations, which is known to be necessary for phonetic discrimination.

Materials and method Participants One hundred and sixty-nine volunteer pregnant mothers (18–40 years old) with healthy uncomplicated pregnancy, carrying singleton fetuses in cephalic position, were recruited at prenatal information sessions at Port-Royal Maternity – CHU Cochin-S. Vincent de Paul University Hospital, in Paris. They were briefed about the experiment and gave written informed consent to participate in the research. Eighty-seven fetuses did not complete testing or their data were rejected for a variety of reasons: They did not enter fetal behavioral state 1F within 2 hours of continuous recording, or their mothers had to leave before they entered state 1F, their behavioral state changed, their baseline HR was under 110 or over 160 bpm (beats per minute), their HR pattern indicated sucking movements (van Woerden, van Geijn, Caron,  2010 Blackwell Publishing Ltd.

van der Valk, Swartjes & Arts, 1988a) or breathing movements (Nijhuis et al., 1982; van Woerden, van Geijn, Swartjes, Caron, Brons & Arts, 1988b) that enhance HR variability, their HR signal was lost for more than 2 s during testing, or their mother had uterine contractions. The final sample consisted of 82 fetuses, 35 males and 47 females. Each fetus was randomly assigned to one of five groups before the test session began and was tested once. Age did not differ across Groups, F(4, 77) = 1.35, p = .26 (see Table 1). All subjects were born at term and were healthy with weights appropriate for gestational age (average weight: 3.429 kg). Stimuli The natural speech sequence (Figure 1a) was an Icelandic sentence from a children’s story, ‘Nanna grtbað mður sna að gefa sr bara eitt tækifæri enn’ (Vanhalewijn & Moerman, 1971). It was spoken by a native Icelandic-speaking woman and recorded with a digital recorder (Sharp MD-MS721H) in a sound-proof room. The Icelandic chimera of the speech sentence was constructed by multiplying the original speech signal with white noise according to the formula yi = xi · randi, where i = time; x = amplitude; rand = random noise (proprietary program by R. Ding). Thus, the chimera had exactly the same temporal structure, i.e. both the fine (fast oscillations) and relatively slow amplitude variations over time, therefore the same rhythm and same attacks as the Icelandic sentence. Its spectral content was that of white noise, thus it lacked all phonological, infrasegmental, information (Figure 1b). The two melodies were played on a piano and then synthesized with MIDI software. Each had nine notes from the same two octave bands, C4 and C5. One had an accelerating ascending melodic contour, from G4 (392 Hz) to B5 (988 Hz; Figure 1c), and the other an accelerating descending melodic contour, from G5 (784 Hz) to E4 (330 Hz; Figure 1d). Although the magnitude of pitch change from one note to the next increased from the beginning to the end of each melody and in opposite directions, the octave-change ratios of successive changes were identical, viz., 1 ⁄ 8, 1 ⁄ 8, 2 ⁄ 8, 1 ⁄ 8, 1 ⁄ 8, 2 ⁄ 8, 3 ⁄ 8 and 2 ⁄ 8. Exaggerating the difference between melodic contours, while keeping their rate of change equal, required some notes between the melodies Table 1 The number of fetuses in each sound condition and their average age in standard deviation (in days) from 38 weeks gestational age (GA) Sound conditions (Groups) Icelandic Icelandic Chimera Descending Melody Ascending Melody Silence All groups

Total Males Females Mean GA: 38 weeks sample N n n SD (in days) 24 22 13 11 12 82

11 9 7 3 5 35

13 13 6 8 7 47

2.87 2.06 2.18 2.62 2.31 2.48

342 Carolyn Granier-Deferre et al.

(a)

(b)

(c)

(d)

Figure 1 Spectrogram (left side) showing the frequency spectrum and sound pressure level variations over time, and envelope (right side) showing the amplitude variations over time, of the Icelandic sentence (a), the Icelandic Chimera (b), the Ascending Melody (c) and the Descending Melody (d).

 2010 Blackwell Publishing Ltd.

Fetuses process temporal features 343

to differ. The total ‘on’ time of the melodies was the same, and the sequences of consecutive eighth notes and quarter notes were identical, giving the melodies the same tempo, rhythm, and thus almost identical amplitude envelopes. In brief, the melodies had four of their nine notes in common, exactly inverse melodic contours, and almost identical amplitude envelopes (see Figure 2). Each stimulus lasted 3.6 s and repeated four times with a 200 ms inter-onset interval. Thus, the total Stimulation Period lasted 15 s. Stimuli were stored on a PC and normalized (Cool Edit Pro software). They were delivered through a loudspeaker with an integrated amplifier and no frequency distortion (Mackie HR624) at an external SPL (ref: 20 lpa) of 90 € 2 dB Leq (Level equivalent: mean SPL during a time period P, which includes silent periods) and 93 dB peak level. Conservatively, their SPL in utero was estimated to be 65–75 dB. All our previous research with sounds at this SPL evokes a high percentage of fetal cardiac decelerations and few accelerations (Lecanuet et al., 1987, 1989, 1992, 1993, 2000). SPL was measured before testing with a sound level meter (‘01dB’ SDB02 model) placed 5 cm in front of the loudspeaker. Background noise in the testing room was 55 € 5 dB. Procedure Upon arrival, mothers sat comfortably in a semi-reclined position on an ergonomically designed chair. We reminded them about the nature of the experiment and gave them a small questionnaire regarding the ongoing quality of the pregnancy, the acoustic character of their home ⁄ work environment, its noise level, type of music they listen to, the languages they speak and hear in daily life, etc. The probe of a Doppler cardiotocograph (Hewlett-Packard M1351A) was positioned on the maternal abdomen to pick up the best quality fetal HR signal. A proprietary software program (R. Humbert & S. Mottet) controlled stimulus presentations and sampled the analog cardiac signal of the cardiotocograph 10 times per second, digitized, and stored each value online. The continuous fetal HR tracing, in bpm, appeared on the PC screen and a reading of the mean and standard deviation of the HR was displayed every 30 s. The loudspeaker was placed 40 cm in front of the mothers’

Figure 2 Score of each of the piano melodies, Ascending and Descending.  2010 Blackwell Publishing Ltd.

abdomen. To mask the fetal stimulus, the mothers wore acoustic noise-canceling headphones (QuietComfort2 BOSE) and heard guitar music at a high, but comfortable level (75–80 dB SPL). Some mothers reported that they sometimes heard a sound but could not identify what it was. Each group was exposed to one of five conditions: A silent control condition, the Icelandic sentence, the Icelandic chimera or one of the two piano melodies. All fetuses were tested in behavioral state 1F, which is considered a fetal equivalent of neonatal quiet sleep. Fetal Heart Rate Pattern A (FHRPA), typical of state 1F, allows observation of low magnitude stimulus-elicited cardiac changes because it is stable, has very low moment-to-moment variability (very narrow oscillation bandwidth), and spontaneous movements and their somato-cardiac components are rare. The analog HR on the computer screen and on the paper tracing of the cardiotocograph (paper speed = 2cm ⁄ mn) was monitored by two experimenters. When both agreed that the fetus was in behavioral state 1F, the mothers put on the headphones. Testing started when the standard deviation of the FHRPA for 5 consecutive minutes was £ 3 bpm. Then, a 60 s Pre-Stimulation Period began. If the HR standard deviation exceeded 3 bpm, the 60 s clock reset. The 15 s Stimulation Period followed immediately. HR recording continued for another 5 minutes, to allow assessment of behavioral state after stimulation stopped. Data analysis First, three experienced observers (CG-D, AR, A-YJ) examined each cardiac paper tracing to verify that the fetus was in 1F, for 5 minutes before, during, and 5 minutes after stimulation. The inclusion criteria were the presence of a continuous FHRPA with a standard deviation no higher than 3 bpm, no or rare movements, as assessed by the fetal movement profile given by the cardiotocograph, and rare spontaneous HR accelerations. The subject was retained only if all three observers agreed. Then each retained data file was checked for missing or erroneous values, which were corrected by linear interpolation. For each subject, HR values, in bpm ⁄ s, were converted to z-scores based on the mean and standard deviation of the 60 s Pre-Stimulation Period. The transformation controls for individual differences in baseline mean HR and HR variability that can affect the amplitude of the stimulus-elicited HR change (Porges, 1974; Lecanuet et al., 1992, 1993). The fetal HR values, in bpm ⁄ s, and their z-score equivalents for each second of the last 15 s of the Pre-Stimulation Period and the 15 s of the Stimulation Period were entered into the statistical software program (Statistica 8). These data were analyzed with ANOVAs having Period (Pre-Stimulation vs. Stimulation) and Time (1 s vs. 2 s vs. 3 s …vs. 15 s) as repeated measures and Group (i.e. Ascending Melody

344 Carolyn Granier-Deferre et al.

vs. Descending Melody vs. Icelandic vs. Chimera vs. Control) as a between-subjects factor. Partial eta-squared (g2p) was used to estimate effect size. It describes the ‘proportion of total variation attributable to the factor, partialling out (excluding) other factors from the total nonerror variation’ (Pierce, Block & Aguinis, 2004, p. 918).

Results Pre-Stimulation Period only A mixed ANOVA on raw values (bpm ⁄ s), with PreStimulation Time (1–15s) as a within-subject factor and Group (Icelandic vs. Chimera vs. Ascending Melody vs. Descending Melody vs. Control) as the between-subjects factor showed no significant effects at all, all F-values were £ 1.04, and had p-values ‡ .39. Therefore, heart rates were very stable and did not differ between Groups during the 15 s before stimulus onset (Table 2). Pre-Stimulation Period vs. Stimulation Period The effect of sound conditions was examined by comparing the second-by-second heart rate in z-scores with a mixed ANOVA, with Period (Pre-Stimulation vs. Stimulation) and Time (1–15s) as within-subject factors and Group as the between-subjects factor (Icelandic vs. Chimera vs. Ascending Melody vs. Descending Melody vs. Control). The main effect of Period, F(1, 77) = 20.83, p < .0001, g2p = 0.21, Time, F(14, 1078) = 9.28, p < .0001, g2p = 0.11, the two-way interactions, Time · Group, F(56, 1078) = 3.26, p < .0001, g2p = 0.14, Period · Time, F(14, 1078) = 6.16, p < .0001, g2p = 0.07, and triple interaction, Period · Time · Group, F(56, 1078) = 1.65, p = .002, g2p = 0.08, were all statistically significant. Figure 3 and Table 3, which show the mean amplitude of the HR change over time, suggest that heart rates decreased with all sound stimuli and that differences between Periods depend on the Group. These differences were examined by subjecting the data of each Group separately to a repeated measure ANOVA with Period and Time as factors. For the Control Group, there were no effects of Period, F(1, 11) = 0,02, p = .89, or Period · Time interaction, F(14, 154) = 0.77, p = .69, but a significant effect of Time, F(14, 154) = 2.18, p = .01, g2p = 0.17. Table 2 Mean (± standard deviation) and range of the fetal heart rate and of the fetal heart rate standard deviation during the Pre-Stimulation Period (N = 82)

Heart rate frequency Heart rate standard deviation

Mean € SD

Range

133.65 bpm € 7.24 1.18 bpm € 0.40

119 bpm to 149 bpm 0.38 bpm to 2.21 bpm

 2010 Blackwell Publishing Ltd.

For each Speech Group, only the main effects of Period and Time were statistically significant, respectively: F(1, 23) = 9.76, p = .005, g2p = 0.30, and F(14, 322) = 1.98, p = .02, g2p = 0.08 for the Icelandic, F(1, 21) = 13.44, p = .001, g2p = 0.39, and F(14, 294) = 2.10, p = .01, g2p = 0.09 for the Chimera. For each Melody Group, only the main effect of Time and Period · Time interaction were significant, respectively: F(14, 168) = 7.63, p < .001, g2p = 0.39 and F(14, 168) = 4.14, p < .001, g2p = 0.26 for the Descending Melody, and F(14, 140) = 6.78, p < .001, g2p = 0.04 and F(14, 140) = 4.21, p < .001, g2p = 0.30 for the Ascending Melody. Interestingly, for the Speech Groups, the effect of Period explained 30% of the variance for the Icelandic Group and 39% for the Chimera Group, while the effect of Time accounted for only 8% of the variance for the Icelandic Group and 9% for the Chimera Group. In contrast, for the Melody Groups, it was the Period · Time interaction which explained 30% of the variance for the Ascending Melody Group and 26% for the Descending Melody Group. Time accounted for even more variance, 40% for the Ascending Melody Group and 39% for the Descending Melody Group. These analyses indicate an important effect of the sound stimulations on the fetal heart rate and that the characteristics of heart rate change after onset of the speech and melodies differed. They also strongly suggest that the category of the stimuli, Music vs. Speech, was the most important factor. As seen in Figure 3 and Table 3, the cardiac decreases differed both in time of occurrence after stimulus onset and in duration. The differences between the patterns of HR decrease as a function of the category of stimulation are examined next. Stimulation Period only We first performed an ANOVA for the Stimulation Period only, without the Control Group, to compare the effects of the different sound stimulations over time, with Group (Icelandic vs. Chimera vs. Ascending Melody vs. Descending Melody) as the between-subjects factor and Time (1–15 s) as the within factor. It showed a main effect of Time, F(14, 924) = 12.81, p < .0001, g2p = 0.16, and a Time · Group interaction, F(42, 924) = 2.33, p < .0001, g2p = 0.10. The interaction confirmed that the different sounds did not affect the fetal HR over time in the same manner; the stimulus effect was time dependent. The Time · Group interaction was further examined with a series of mixed ANOVAs, with Group as a between-subjects factor and Time (1–15 s) as a within factor. Melody Groups Two separate ANOVAs showed that each Melody Group was statistically different from the Control Group tested in silence. There was a main effect of Time, and the

Fetuses process temporal features 345

Figure 3 Mean Heart Rate of the Silent Control (plain line), Icelandic (open circles), Chimera (filled squares), Ascending Melody (large broken lines) and Descending Melody (small broken lines) Groups, during the last 15 s of the Pre-Stimulation Period and the following 15 s Stimulation Period, in z-scores; z-score values were based on the mean and standard deviation of each subject’s Heart Rate during the 60 s Pre-Stimulation Period.

Group · Time interaction was significant for the Descending Melody Group, respectively, F(14, 322) = 4.18, p < .0001, g2p = 0.15, F(14, 322) = 6.45, p < .0001, g2p = 0.22, and for the Ascending Melody Group, respectively, F(14, 294) = 3.82, p < .0001, g2p = 0.15, F(14, 294) = 8.21, p < .0001, g2p = 0.28. When the Ascending and Descending Melody Groups were compared, only the main effect of Time was statistically significant, F(14, 308) = 11.30, p < .0001, g2p = 0.34. There was no effect of Group, F(1, 22) = 0.06, p = .81, or Group · Time interaction, F(14, 308) = 0.77, p = .70. Both melodies induced a monophasic HR deceleration that peaked 6–7 s after stimulus onset and returned to baseline 13 s after stimulus onset. For the Descending Melody the peak, z = )3.29 bpm, occurred 6 s after onset, for the Ascending Melody the peak, z = )2.68 bpm, occurred 7 s after onset. There was no difference in peak deceleration values either 6 s, F(1, 22) = 0.40, p = .53, or 7 s after onset of stimulation, F(1, 22) = 0.15, p = .70. Since the patterns of the cardiac response did not differ between the two Melody Groups, their data were pooled in subsequent analyses.

Speech Groups Each Speech Group differed from the Control Group tested in silence. When the Icelandic and Control Groups were compared, there was a statistically significant effect of Group, F(1, 34) = 4.35, p = .04, g2p = 0.11, and Group · Time interaction, F(14, 476) = 2.18, p = .008, g2p = 0.06. The Chimera and Control Groups differed as well, with a Group, F(1, 32) = 6.77, p = .01, g2p = 0.17, and a Group · Time interaction, F(14, 448) = 1.76, p = .04, g2p = 0.05. When the Icelandic and Chimera Groups were compared, only the main effect of Time was statistically significant, F(14, 616) = 3.57, p < .0001, g2p = 0.07. There was no effect of Group, F(1, 44) = 1.37, p = .25, or Group · Time interaction, F(14, 616) = 0.33, p = .99. The cardiac decelerations elicited by both stimuli were very similar. Figure 3 shows that the heart rate response to each speech stimulus appears to be multiphasic with two deceleration peaks, an early peak at 4 s after stimulus onset and another 10 s after, respectively, z = )2.04 and z = )2.05 for the Chimera, z = )1.22 and z = )1.77 for the Icelandic. The two

Table 3 Amplitude (mean z-scores ± standard error of the mean) of the maximum heart rate change from prestimulus level, z = 0, and its latency during the 15 s of the Stimulation Period Sound conditions (Groups) N = 82

Silence (Control gr.) n = 12

Ascending Melody n = 11

Descending Melody n = 13

Pooled Melody n = 24

Icelandic n = 24

Icelandic Chimera N = 22

Peak amplitude € SEM Latency of peak amplitude

+0.67 € 0.82

)2.68 € 0.81

)3.29 € 1.16

)2.99 € 0.73

)1.77 € 0.65

)2.05 € 0.57

7s

6s

7s

10 s

10 s

9s

 2010 Blackwell Publishing Ltd.

346 Carolyn Granier-Deferre et al.

peak deceleration values did not differ between the Icelandic and the Chimera groups, either 4 s after stimulus onset, F(1, 44) = 1.70, p = .20, or 10 s after, F(1, 44) = 0.09, p = .76. Furthermore, the heart rates did not return to baseline within 15 s of stimulus onset: The deceleration elicited by the Icelandic returned to baseline level 18 s after onset and the one elicited by the Chimera 17 s after onset. The analyses indicate that the heart rate decelerations of the Speech Groups did not differ in pattern or peak amplitudes. The effect sizes observed here, when each experimental group was compared to the Control Group, confirmed that the category of stimulation per se had the single greatest effect on heart rate. For the Melody Groups, the Group · Time interaction explained most of the variance, 22% with the Descending Melody and 28% with the Ascending Melody. For the Speech Groups it was the main effect of Group that explained most of the variance, 11% with the Icelandic Group and 17% with the Chimera Group. Essentially no variance arose from differential effects of the Ascending vs. Descending Melodies or from differential effects of Icelandic Speech vs. Chimera. Comparison of the Melody and Speech Groups A mixed ANOVA was used to compare the pooled Melody, Icelandic, and Chimera Groups. Time during the 15 s Stimulation Period was the within-subjects factor. There was a main effect of Time, F(14, 938) = 10.12, p < .0001, g2p = 0.13, and a Group · Time interaction, F(28, 938) = 3.09, p < .0001, g2p = 0.08. Separate ANOVAs comparing each Speech Group to the pooled Melody Group confirmed that the cardiac changes to the Icelandic and the Chimera Groups were significantly different from that of the pooled Melody Group. With the Icelandic sentence, there was a significant effect of Time, F(14, 644) = 10.56, p < .0001, g2p = 0.19, and a Group · Time interaction, F(14, 644) = 5.24, p < .0001, g2p = 0.10. With the Chimera there was also a main effect of Time, F(14, 616) = 9.11, p < .0001, g2p = 0.17, and a Group · Time interaction, F(14, 616) = 3.63, p < .0001, g2p = 0.08. The peak response observed in the pooled Melody Group had a greater amplitude, z = )2.99, than the peak response to the Icelandic, z = )1.77, and to the Chimera, z = )2.05. It also peaked earlier, 7 s after stimulus onset, than did the response to Icelandic, 10 s, and to the Chimera, 10 s, and it returned to baseline within 12 s of stimulus onset while the responses to the speech stimuli did not reach baseline values within the 15 s of stimulation. Therefore, the cardiac deceleration elicited by the pooled melodies was greater, peaked earlier, and returned to baseline earlier than did the cardiac decelerations to the Icelandic and to the Chimera, which were more shallow and sustained, and did not differ from each other.  2010 Blackwell Publishing Ltd.

Discussion The perception of speech and music melodies requires processing of variations in spectra and amplitude over different temporal windows. A review of the literature on the characteristics of sounds available to the fetal ear, the functional development of the fetal auditory system, fetal auditory perception and the newborn data related to prenatal familiarization to music and speech stimuli indicated that the fetal auditory system processes, and the fetus, can perceive variations in frequency and spectra, and should be able to process variations in amplitude as well. The study investigated fetal auditory processing of variations in amplitude and spectra with four auditory streams lasting 15 s: Two piano melodies, containing only relatively slow variations in amplitude and spectra over time, an Icelandic sentence containing rapid and relatively slow amplitude and spectral variations over time, and a sound chimera of the Icelandic sentence, composed of broadband noise and containing only the temporal variations in amplitude of the original sentence. We found that all stimulated groups showed a statistically significant HR decrease starting shortly after stimulus onset and differed significantly from the silent control group (see Figure 3). Most significantly, the response patterns to the four sound streams differed categorically. The cardiac decelerations with the piano melodies differed significantly in their time course and magnitude from the decelerations with the speech stimuli. HR responses to the piano melodies were the same. Both elicited a brief monophasic HR deceleration that peaked at 3 SD below pre-stimulus level 6–7 s after stimulus onset, and then returned to baseline within 5 s. The piano melodies differed in spectral content and contour but both were composed of the same sequence of eighth notes and quarter notes that unfolded over time with the same tempo and rhythm. Hence, they had almost identical patterns of slow amplitude variations within notes and over time. Interestingly, the pattern of response to the melodies is not unique to this particular experiment. It has been observed in other near-term fetal studies with sound streams containing no amplitude variation, or only relatively slow spectral and amplitude variations. This monophasic response was found with longer versions of the piano melodies used here delivered at 80 dB Leq (Granier-Deferre, Bassereau, Jacquet & Lecanuet, 1998), with sequences of repeating piano notes, D4 or C5, presented at 93 dB SPL (Lecanuet et al., 2000), and also with 5 s octave-band noises, centered at 500 Hz, 2000 Hz and 5000 Hz, delivered at 100 dB SPL (Lecanuet et al., 1988). Moreover, a figure in Groome, Loizou, Holland, Law, Mooney and Dykman (2000) shows that the deceleration elicited by an 8 s emission of the vowel ⁄ i ⁄ followed by 7 s of ⁄ æ ⁄ at 83 dB peaked 7 s after onset of the sequence and returned to baseline 5 s later, exactly like the ones

Fetuses process temporal features 347

elicited by our melodies. All these decelerations have the characteristics of the ‘transient-detecting-reflex’ described in newborns by Graham (1992), i.e. a brief short latency HR deceleration that occurs at the onset of the stimulus. They are elicited in sleep and waking states as well. For Richards and Casey (1991, 1992; Richards, 1997) they mean ‘stimulus orienting’ in the awake infant. HR responses to the speech streams were the same. Both evoked a sustained decrease with two deceleration peaks, whose maximum magnitude of about 2 SD below pre-stimulus level occurred 10 s after stimulus onset, and had not yet returned to baseline before the 15 s stimulus ended. The Icelandic sentence and its Chimera had vastly different spectral characteristics but they had exactly the same sequence of concatenated rapid and slow temporal amplitude variations over time. The longer duration of the fetal HR deceleration to the speech streams is robust. It occurred in an unpublished pilot study with 11 fetuses, where the stimulation was a series of different English sentences, from the corpus of T. Nazzi (see, Nazzi, Bertoncini & Mehler, 1998). That speech stream elicited a deep sustained deceleration that peaked at 3 SD below pre-stimulus level 10 s after stimulus onset, and the HR was still 2 SD below pre-stimulus level when the 15 s stream ended. The difference in magnitude of the HR decelerations during the Icelandic and English streams is probably due to the much larger SPL variations within the English stream compared to the Icelandic stream: the English stream was delivered at 90 dB Leq, like the Icelandic and the chimera, but had a 100 dB peak level. This sustained HR deceleration was also recorded with other speech stimuli in the near-term fetus, a repeating paired syllables [bi]–[ba] and [ba]–[bi] (Lecanuet et al., 1987), and with a repeating natural short French sentence (Lecanuet et al., 1992, 1993). Interestingly, Fifer and Moon (1995) also reported late HR decelerations to adult-directed speech spoken by the fetus’ mothers and other females. These HR responses fit the ‘generalized orientation reflex’ described by Graham (1992), a sustained, long latency decrease. It is often considered specific to the awake infant but has also been found in sleeping neonates. It is usually interpreted as the correlate of enhanced sensitivity to the stimulus or stimulus discrimination. However, the prolonged decelerations that we recorded during the speech streams more clearly resemble the ‘heart-rate-defined sustained attention’ described in infants by Richards and Casey (1991, 1992) and Richards (1997). Sustained attention is characterized by a sustained lowered heart rate that immediately follows the ‘stimulus orienting’ phase; ‘attention termination’ occurs when the heart rate returns to pre-stimulus level. This sustained cardiac response is associated with more focused information processing in the infant. The multiphasic pattern of the heart rate decrease observed in the two speech groups could reflect the two successive phases described by Richards and Casey (1991, 1992). It might also indicate that fetal attention was first engaged  2010 Blackwell Publishing Ltd.

and then re-engaged by the repetition of the Icelandic sentence and the chimera because of the presence of rapid transitions. The melodies have a simple temporal structure and they elicited a monophasic response. Consequently, the categorical differences in music melodies and speech map onto categorical differences in parasympathic responses, with music melodies eliciting a simple detecting ⁄ orienting response, and speech streams eliciting a sustained orientation or attentional response that lasted as long as the streams themselves. A conservative interpretation of the HR data is that the near-term fetus can perceive sound streams, and not merely react to the onset of a sound stimulation, and process the music melodies and speech streams differently. HR change during the two melodies did not differ, despite differences in their spectral content and extreme differences in the direction, ascending vs. descending, of spectral change over time. HR change during the Icelandic sentence and its chimera did not differ, despite extreme differences in their spectral content (see Figure 3). Therefore, differences in spectra or spectral variations over time did not elicit any difference in fetal response patterns to either the melodies or speech streams. The categorical difference in HR response can be attributed solely to the major categorical difference between the melodies and the speech streams, specifically, the rapid temporal variations in amplitude present only in the speech streams. Our results indicate that the near-term fetus can process the rapid temporal variations in amplitude, of the order of tens of ms, that are specific to speech and necessary for phonetic discriminations. They also show that this processing can occur in utero independently of any spectral cues. The temporal variations in speech, such as the rhythmic properties of speech and language, are considered to play a major role in the development of speech perception and language discrimination in infancy (Bertoncini, Floccia, Nazzi & Mehler, 1995; Nazzi et al., 1998; Ramus, Hauser, Miller, Morris & Mehler, 2000). The present results indicate that these properties too can be processed before birth in the absence of frequency variations or spectral cues, such as those carried by vowels. Our conclusions follow reasonably from the data but do not imply that spectral processing does not occur. On the contrary, for example, we have shown that by changing the procedure of the present study to one where the chimera is followed immediately by the sentence, and vice versa, the stimulus change elicits a new cardiac deceleration. This indicates that fetuses detected the difference in the spectral characteristics of the two stimuli. We have replicated this result with newborns (Ribeiro, Granier-Deferre & Jacquet, in preparation). Thus, both naturally associated aspects of a speech stream are processed, but research has tended to examine, intentionally or not, one process at a time. In brief, we showed that near-term fetuses can perceive sound streams, process the temporal variations in amplitude specific to speech sounds, and display different levels of attentional neuro-vegetative responses to a

348 Carolyn Granier-Deferre et al.

stream that vary according to its temporal complexity. The body of research reviewed in the introduction showed that they can discriminate spectra as well. The temporal and spectral properties of a speech stream, which are naturally associated in normal speech, can be processed in utero. Therefore, we propose that these auditory capacities allow the development of speechsound perception per se, including phonetic discriminations, to begin before birth with repeated exposure to the spectral and non-random temporal characteristics of the maternal voice, speech, and language and, as absolute auditory thresholds decrease, to the spectral and temporal characteristics of the voice, speech, and language of others as well. These fetal auditory capacities can explain how newborns can recognize their mother’s voice independently of what she is saying (e.g. DeCasper & Fifer, 1980), i.e. independently of the specific temporal variations in her speech, and also recognize a specific passage she often recited before birth, independently of who recites it after birth (DeCasper & Spence, 1986), i.e. independently of the specific spectral properties of her voice.

Acknowledgements We express our gratitude to the mothers and to the babies involved in this study and to Professor Cabrol, head of the Port-Royal Maternity, and his staff, for welcoming our group. We thank Professor A.J. DeCasper for his helpful comments and English editing of the manuscript, Pr. C. Grard for devising the music stimuli, R. Ding for devising the Icelandic sound chimera, Dr T. Nazzi for allowing us to use his corpus of foreign language sentences in our pilot study, and R. Humbert and S. Mottet for the cardiac data acquisition software. Thanks also to Dr Asta Georgsdottir for recording the Icelandic sentence. This article is dedicated to our dear friend and colleague Jean-Pierre Lecanuet, who died suddenly in 2002. He was a pioneering figure in the study of fetal perception in all sensory modalities. He had become interested in fetal perception of the temporal aspects of sound and completed a study on tempo discrimination. Sadly, he had not completed the manuscript before he passed away.

References Abrams, R.M., Gerhardt, K.J., & Peters, A.J.M. (1995). Transmission of sound and vibration to the fetus. In J.-P. Lecanuet, W.P. Fifer, N.A. Krasnegor, & W.P. Smotherman (Eds.), Fetal development: A psychobiological perspective (pp. 315–330). Hillsdale, NJ: Lawrence Erlbaum Associates. Abrams, R.M., Griffiths, S.K., Huang, X., Sain, J., Langford, G., & Gerhardt, K.J. (1998). Fetal music perception: the role of sound transmission. Music Perception, 15, 307–317. Armitage, S.E., Baldwin, B.A., & Vince, M.A. (1980). The fetal sound environment of sheep. Science, 208, 1173–1174.

 2010 Blackwell Publishing Ltd.

Benzaquen, S., Gagnon, R., Hunse, C., & Foreman, J. (1990). The intrauterine sound environment of the human fetus during labor. American Journal of Obstetrics and Gynecology, 163, 484–490. Berg, W.K., & Berg, K.M. (1987). Psychophysiological development in infancy: state, sensory function and attention. In J.D. Osofsky (Ed.), Handbook of infant development (pp. 238– 317). New York: Wiley and Sons. Bertoncini, J., Floccia, C., Nazzi, T., & Mehler, J. (1995). Morae and syllables: rythmical basis of speech representations in neonates. Language and Speech, 38, 311–329. Bertoncini, J., Serniclaes, W., & Lorenzi, C. (2009). Discrimination of speech sounds based upon temporal envelope versus fine structure cues in 5- to 7-year-old children. Journal of Speech, Language, and Hearing Research, 52, 682–695. Birnholz, J.C., & Benacerraf, B.B. (1983). The development of human fetal hearing. Science, 222, 516–518. Busnel, M.-C., & Granier-Deferre, C. (1983). And what of fetal audition? In A. Oliverio & M. Zapella (Eds.), The behavior of human infants (pp. 93–126). New York: Plenum Press. Busnel, M.-C., Granier-Deferre, C., & Lecanuet, J.-P. (1992). Fetal audition. Annals of the New York Academy of Sciences, 662, 118–134. Clarkson, M.G., & Berg, W.K. (1983). Cardiac orienting and vowel discrimination in newborns: crucial stimulus parameters. Child Development, 54, 162–171. Cooper, R.P., & Aslin, R.N. (1989). The language environment of the young infant: implications for early perceptual development. Canadian Journal of Psychology, 43, 247–265. DeCasper, A.J., & Fifer, W.P. (1980). Of human bonding: newborns prefer their mother’s voice. Science, 208, 1174– 1176. DeCasper, A.J., Lecanuet, J.-P., Busnel, M.-C., Granier-Deferre, C., & Maugeais, R. (1994). Fetal reaction to recurrent maternal speech. Infant Behavior and Development, 17, 159– 164. DeCasper, A.J., & Prescott, P. (2009). Lateralized processes constrain auditory reinforcement in human newborns. Hearing Research, 255, 135–141. DeCasper, A.J., & Spence, M.J. (1986). Prenatal maternal speech influences newborns’ perception of speech sounds. Infant Behavior and Development, 9, 133–150. DeCasper, A.J., & Spence, M.J. (1991). Auditorially mediated behavior during the perinatal period: a cognitive view. In M.J. Weiss & P.R. Zelazo (Eds.), Infant attention (pp. 142– 176). Norwood, NJ: Ablex. Dehaene-Lambertz, G., Hertz-Pannier, L., & Dubois, J. (2006). Nature and nurture in language acquisition: anatomical and functional brain-imaging studies in infants. Trends in Neurosciences, 29, 367–373. Draganova, R., Eswaran, H., Murphy, P., Huotilainen, M., Lowery, C., & Preissl, H. (2005). Sound frequency change detection in fetuses and newborns: a magnetoencephalographic study. NeuroImage, 28, 354–361. Draganova, R., Eswaran, H., Murphy, P., Lowery, C., & Preissl, H. (2007). Serial magnetoencephalographic study of fetal and newborn auditory discriminative evoked responses. Early Human Development, 83, 199–207. Eldredge, L., & Salamy, A. (1996). Functional auditory development in preterm and full term infants. Early Human Development, 45, 215–228.

Fetuses process temporal features 349

Eswaran, H., Lowery, C.L., Robinson, S.E., Wilson, J.D., Cheyne, D., & McKenzie, D. (2000). Challenges of recording human fetal auditory evoked response using magnetoencephalography. Journal of Maternal-Fetal Medicine, 9, 303–307. Fifer, W.P., & Moon, C.M. (1994). The role of mother’s voice in the organization of brain function in the newborn. Acta Paediatrica, 397, 86–93. Fifer, W.P., & Moon, C.M. (1995). Effects of fetal experience with sound, In J.-P. Lecanuet, W.P. Fifer, N.A. Krasnegor, & W.P. Smotherman (Eds.), Fetal development: A psychobiological perspective (pp. 351–366). Hillsdale, NJ: Lawrence Erlbaum Associates. Forbes, H.S., & Forbes, H.B. (1927). Fetal sense reaction: hearing. Journal of Comparative Physiological Psychology, 7, 353–355. Fulford, J., Vadeyar, S.H., Dodampahala, S.H., Ong, S., Moore, R.J., Baker, P.N., James, D.K., & Gowland, P. (2004). Fetal brain activity and hemodynamic response to a vibroacoustic stimulus. Human Brain Mapping, 22, 116–121. Gagnon, R., Benzaquen, S., & Hunse, C. (1992). The fetal sound environment during vibroacoustic stimulation in labor: effect on fetal heart rate response. Obstetrics and Gynecology, 79, 950–955. Gerhardt, K.J. (1989). Characteristics of the fetal sheep sound environment. Seminars in Perinatology, 13, 362–370. Gerhardt, K.J., Huang, X., Arrington, K.E., Meixner, K., Abrams, R.M., & Antonelli, P J. (1996). Fetal sheep in utero hear through bone conduction. American Journal of Otolaryngology, 17, 374–379. Gerhardt, K.J., Otto, R., Abrams, R.M., Colle, J.J., Burchfield, D.J., & Peters, A.J. (1992). Cochlear microphonics recorded from fetal and newborn sheep. American Journal of Otolaryngology, 13, 226–233. Graham, F.K. (1992). Attention: the heart beat, the blink, and the brain. In B.A. Campbell, H. Hayne, & R. Richardson (Eds.), Attention and information processing in infants and adults: Perspectives from human and animal research (pp. 3– 29). Hillsdale, NJ: Erlbaum. Graham, F.K., Anthony, B.J., & Zeigler, B.L. (1983). The orienting response and developmental processes, In D. Siddle (Ed.), Orienting and habituation: Perspectives in human research (pp. 340–371). Chichester: John Wiley. Granier-Deferre, C., Bassereau, S., Jacquet, A.-Y., & Lecanuet, J.-P. (1998). Fetal and neonatal cardiac orienting response to music in quiet sleep. Developmental Psychobiology, 33, 372. Granier-Deferre, C., & Lecanuet, J-.P. (1987). Influence de stimulations auditives prcoces sur la maturation anatomique et fonctionnelle du systme auditif. Progrs en Nonatologie, 7, 236–249. Granier-Deferre, C., Lecanuet, J.-P., Cohen, H., & Busnel, M.-C. (1985). Feasibility of prenatal hearing test. Acta Otolaryngol (Stockholm), 421, 93–101. Granier-Deferre, C., Lecanuet, J.-P., Cohen, H., Busnel, M.-C., Querleu, D., & Sureau, C. (1984). Le foetus entend. Les dossiers de l’obsttrique, 107, 27–31. Griffiths, S.J., Brown, W.S. Jr., Gerhardt, K.J., Abrams, R.M., & Morris, R.J. (1994). The perception of speech sounds recorded within the uterus of a pregnant sheep. Journal of the Acoustical Society of America, 96, 2055–2063. Groome, L.J., Loizou, P.C., Holland, S.B., Law, D.J., Mooney, D.M., & Dykman, R.A. (2000). Fetuses and neonates have different heart rate responses to low-intensity stimulation in quiet sleep. Infant Behavior and Development, 23, 61–77.  2010 Blackwell Publishing Ltd.

Groome, L.J., Mooney, D.M., Holland, S.B., Bentz, L.S., & Atterbury, J.L. (1997). The heart rate deceleratory response in low-risk human fetuses: effect of stimulus intensity on response topography. Developmental Psychobiology, 30, 106– 113. Groome, L.J., Mooney, D.M., Holland, S.B., Smith, L.A., Atterbury, J.L., & Dykman, R.A. (1999). Behavioral state affects heart rate response to low-intensity sound in human fetuses. Early Human Development, 54, 39–54. Hepper, P.G. (1988). Fetal ‘soap’ addiction. Lancet, 1, 1147– 1148. Hepper, P.G. (1991). An examination of fetal learning before and after birth. Irish Journal of Psychology, 12, 95–107. Holst, M., Eswaran, H., Lowery, C., Murphy, P., Norton, J., & Preissl, H. (2005). Development of auditory evoked fields in human fetuses and newborns: a longitudinal MEG study. Clinical Neurophysiology, 116, 1949–1955. Horner, K., Servire, J., & Granier-Deferre, C. (1987). Deoxyglucose demonstration of in utero hearing in the guineapig foetus. Hearing Research, 26, 327–333. Huang, X., Gerhardt, K.J., Griffiths, S.K., & Abrams, R.M. (2002). Intelligibility of speech recorded from the cochlea of fetal sheep in utero. Abstract Midwinter Res. Meet. Assoc. Res. Otolaryngology, 25, 52. Huotilainen, M., Kujala, A., Hotakainen, M., Parkkonen, L., Taulu, S., Simola, J., Nenonen, J., Karjalainen, M., & Ntnen, R. (2005). Short-term memory functions of the human fetus recorded with magnetoencephalography. NeuroReport, 16, 81–84. Jardri, R., Pins, D., Houfflin-Debarge, V., Chaffiotte, C., Rocourt, N., Pruvo, J.-P., Steinling, M., Delion, P., & Thomas, P. (2008). Fetal cortical activation to sound at 33 weeks of gestation: a functional MRI study. NeuroImage, 42, 10–18. Kisilevsky, B.S. (1995). The influence of stimulus and subject variables on human fetal responses to sound and vibration. In J.-P. Lecanuet, W.P. Fifer, N.A. Krasnegor, & W.P. Smotherman (Eds.), Fetal development: A psychobiological perspective (pp. 263–278). Hillsdale, NJ: Lawrence Erlbaum Associates. Kisilevsky, B.S., Hains, S.M.J., Jacquet, A.-Y., Granier-Deferre, C., & Lecanuet, J.-P. (2004). Maturation of fetal perception of music. Developmental Science, 7, 550–559. Kisilevsky, B.S., Hains, S.M.J., Lee, K., Xie, X., Huang, H., Ye, H.-H., Zhang, K., & Wang, Z. (2003). Effects of experience on fetal voice recognition. Psychological Science, 14, 220–224. Kisilevsky, B.S., & Low, J.A. (1998). Human fetal behavior: 100 years of study. Developmental Review, 18, 1–29. Kisilevsky, B.S., Muir, D.W., & Low, J.A. (1992). Maturation of human fetal responses to vibroacoustic stimulation. Child Development, 63, 1497–1508. Krueger, C., Holditch-Davis, D., Quint, S., & DeCasper, A.J. (2004). Recurring auditory experience in the 28- to 34-weekold fetus. Infant Behavior and Development, 27, 537–543. Krumholz, A., Felix, J.K., Goldstein, P.J., & McKenzie, E. (1985). Maturation of the brain-stem auditory evoked potential in premature infants. Electroencephalography and Clinical Neurophysiology, 62, 124–134. Kuhl, P.K. (1991). Human adults and human infants show a ‘perceptual magnet effect’ for the prototypes of speech categories, monkeys do not. Perception & Psychophysics, 50, 93– 107.

350 Carolyn Granier-Deferre et al.

Kuhl, P.K. (2000). A new view of language acquisition. Proceedings of the National Academy of Sciences, USA, 97, 11850–11857. Kuhl, P.K., Conboy, B.T., Coffey-Corina, S., Padden, D., Rivera-Gaxiola, M., & Nelson, T. (2008). Phonetic learning as a pathway to language: new data and native language magnet theory expanded (NLM-e). Philosophical Transactions of the Royal Society B, 363, 979–1000. Lecanuet, J.-P., Gautheron, B., Locatelli, A., Schaal, B., Jacquet, A.-Y., & Busnel, M.-C. (1998). What sounds reach fetuses: biological and non-biological modeling of the transmission of pure tones. Developmental Psychobiology, 33, 203–220. Lecanuet, J.-P., & Granier-Deferre, C. (1993). Speech stimuli in the fetal environment. In B. de Boysson-Bardies, S. de Schoenen, P. Jusczyck, P. McNeilage, & J. Morton (Eds.), Developmental neurocognition: Speech and face processing in the first year of life, NATO, ASI Series (Series D), Vol. 69 (pp. 237–248). Dordrecht, Boston, London: Kluwer Academic Publisher. Lecanuet, J.-P., Granier-Deferre, C., & Busnel, M.-C. (1988). Fetal cardiac and motor responses to octave-band noises as a function of central frequency, intensity and heart rate variability. Early Human Development, 18, 81–93. Lecanuet, J.-P., Granier-Deferre, C., & Busnel, M.-C. (1989). Differential fetal auditory reactiveness as a function of stimulus characteristics and state. Seminars in Perinatology, 13, 421–429. Lecanuet, J.-P., Granier-Deferre, C., & Busnel, M.-C. (1991). Prenatal familiarization. In G. Piraut-Le Bonniec & M. Dolitsky (Eds.), Language bases discourse bases (pp. 31–44). Amsterdam and Philadelphia, PA: John Benjamin. Lecanuet, J.-P., Granier-Deferre, C., & Busnel, M.-C. (1995). Human fetal auditory perception. In J.P. Lecanuet, W.P. Fifer, N.A. Krasnegor, & W.P. Smotherman (Eds.), Fetal development: A psychobiological perspective (pp. 239–262). Hillsdale, NJ: Lawrence Erlbaum Associates. Lecanuet, J.-P., Granier-Deferre, C., Cohen, H., Le Houezec, R., & Busnel, M.-C. (1986). Fetal responses to acoustic stimulation depend on heart rate variability pattern, stimulus intensity, and repetition. Early Human Development, 13, 269–283. Lecanuet, J.-P., Granier-Deferre, C., DeCasper, A.J., Maugeais, R., Andrieu, A.J., & Busnel, M.-C. (1987). Perception et discrimination foetales de stimuli langagiers; mise en vidence partir de la ractivit cardiaque, rsultats prliminaires. Comptes-Rendus de l’Acadmie des Sciences, Paris, Srie III, 305, 161–164. Lecanuet, J.-P., Granier-Deferre, C., Jacquet, A.-Y., & Busnel, M.-C. (1992). Decelerative cardiac responsiveness to acoustical stimulation in the near term fetus. Quarterly Journal of Experimental Psychology, 44b, 279–303. Lecanuet, J.-P., Granier-Deferre, C., Jacquet, A.-Y., Capponi, I., & Ledru, L. (1993). Prenatal discrimination of a male and a female voice uttering the same sentence. Early Development and Parenting, 2, 217–228. Lecanuet, J.-P., Granier-Deferre, C., Jacquet, A.-Y., & DeCasper, A.J. (2000). Fetal discrimination of low-pitched musical notes. Developmental Psychobiology, 36, 29–39. Lecanuet, J.-P., Jacquet, A.-Y., & Bontemps, A. (2002). Discrimination of fast isochronous sequences by end of gestation fetuses. Unpublished manuscript. (See acknowledgements.)

 2010 Blackwell Publishing Ltd.

Lorenzi, C., Dumont, A., & Fllgrabe, C. (2000). Use of temporal envelope cues by children with developmental dyslexia. Journal of Speech, Language, and Hearing Research, 43, 1367–1379. Mastropieri, D., & Turkewitz, G. (1999). Prenatal experience and neonatal responsiveness to vocal expressions of emotion. Developmental Psychobiology, 35, 204–214. Mirman, D., Holt, L., & McClelland, J.L. (2004). Categorization and discrimination of nonspeech sounds: differences between steady state and rapidly changing acoustic cues. Journal of the Acoustical Society of America, 116, 1198–1207. Moon, C., Cooper, R., & Fifer, W.P. (1993). Two-day-olds prefer their native language. Infant Behavior and Development, 16, 495–500. Moon, C., & Fifer, W.P. (2000). Evidence of transnatal auditory learning. Journal of Perinatology, 20, 37–44. Moore, D.R. (2002). Auditory development and the role of experience. British Medical Bulletin, 63, 171–181. Moore, J.K. (2002). Maturation of human auditory cortex: implications for speech perception. Annals of Otology, Rhinology, & Laryngology – Supplement, 189, 7–10. Moore, J.K., Ponton, C.W., Eggermont, J., Wu, B.J.-C., & Huang, J.Q. (1996). Perinatal maturation of the auditory brain stem response: changes in path length and conduction velocity. Ear and Hearing, 17, 411–418. Moore, R.J., Vadeyar, S., Fulford, J., Tyler, D. J., Gribben, C., Baker, P.N., James, D.K., & Gowland, P.A. (2001). Antenatal determination of fetal brain activity in response to an acoustic stimulus using functional magnetic resonance imaging. Human Brain Mapping, 12, 94–99. Morlet, T., Collet, L., Salle, B., & Morgon, A. (1993). Functional maturation of cochlear active mechanisms and of the medial olivocochlear system in humans. Acta Otolaryngology (Stockholm), 113, 271–277. Morlet, T., Lapillonne, A., Ferber, C., Duclaux, R., Sann, L., Putet, G., Salle, B., & Collet, L. (1995). Spontaneous otoacoustic emissions in preterm neonates: prevalence and gender effects. Hearing Research, 90, 44–54. Nazzi, T., Bertoncini, J., & Mehler, J. (1998). Language discrimination by newborns: towards an understanding of the role of rhythm. Journal of Experimental Psychology: Human Perception and Performance, 24, 756–766. Nijhuis, J.G., Prechtl, H.F.R., Martin, C.B., & Bots, R.S.G.M. (1982). Are there behavioural states in the human fetus? Early Human Development, 6, 177–195. Oliver, C.C. (1989). Sound and vibration transmission in tissues. Seminars in Perinatology, 13, 354–361. Pasman, R.L., Ntanen, R., & Alho, K. (1991). Auditory evoked responses in prematures. Infant Behavior and Development, 14, 129–135. Peiper, A. (1925). Sinnesempfindugen des Kinder vor Seiner Geburt. Monatsschrift Kinderheilkunde, 29, 236–241. Peters, A.J.M., Abrams, R.M., Gerhardt, K.J., & Griffiths, S.K. (1993). Transmission of airborne sounds from 50– 20,000 Hz into the abdomen of sheep. Journal of Low Frequency Noise & Vibration, 12, 16–24. Petitjean, C. (1989). Une condition de l’audition foetale: la conduction sonore osseuse. Consquences cliniques et applications pratiques. MD dissertation, University of Franche-Comt, BesanÅon (available from the University library).

Fetuses process temporal features 351

Pierce, C.A., Block, R.A., & Aguinis, H. (2004). Cautionary note on reporting eta-squared values from multifactor anova designs. Educational and Psychological Measurement, 64, 916–924. Pierson, L.L., Gerhardt, K.J., Griffiths, S.K., & Abrams, R.M. (1995). Auditory brainstem response in sheep. Part I: Fetal development. Developmental Psychobiology, 28, 293– 305. Poeppel, D. (2003). The analysis of speech in different temporal integration windows: cerebral lateralization as asymmetric sampling in time. Speech Communication, 41, 245–255. Porcaro, C., Zappasodi, F., Barbati, G., Salustri, C., Pizzella, V., Rossini, P.M., & Tecchio, F. (2006). Fetal auditory responses to external sounds and mother’s heart beat: detection improved by Independent Component Analysis. Brain Research, 1101, 51–58. Porges, S.W. (1974). Heart rate indices of newborn attentional responsiveness. Merrill-Palmer Quarterly, 20, 231–253. Pujol, R., Laville-Rebillard, M., & Lenoir, M. (1998). Development of sensory and neural structures in the mammalian cochlea. In E.W. Rubel, A.N. Popper, & R.R. Fay (Eds.), Development of the auditory system. Springer handbook of auditory research (pp. 146–193). New York: SpringerVerlag. Pujol, R., Lavigne-Rebillard, M., & Uziel, A. (1991). Development of the human cochlea. Acta Otolarnygology (Stockholm), Suppl 482, 7–12. Pujol, R., & Uziel, A. (1988). Auditory development: peripheral aspects. In P.F. Timiras & E. Meisami (Eds.), Handbook of human biologic development (pp. 109–130). Boca Raton, FL: CRC Press. Querleu, D., Renard, X., Boutteville, C., & Crpin, G. (1989). Hearing by the human fetus? Seminars in Perinatology, 13, 409–420. Querleu, D., Renard, X., & Crpin, G. (1981). Perception auditive et ractivit fœtale aux stimulations sonores. Journal de Gyncologie Obsttrique et de Biologie de la Reproduction, 10, 307–314. Querleu, D., Renard, X., Versyp, F., Paris-Delrue, L., & Crpin, G. (1988a). Fetal hearing. European Journal of Obstetrics & Gynecology and Reproductive Biology, 29, 191–212. Querleu, D., Renard, X., Versyp, F., Paris-Delrue, L., & Vervoot, P. (1988b). La transmission intra-amniotique des voix humaines. Revue FranÅaise de Gyncologie et Obsttrique, 83, 43–50. Ramus, F., Hauser, M.D., Miller, C., Morris, D., & Mehler, J. (2000). Language discrimination by human newborns and by cotton-top tamarin monkeys. Science, 288, 349–351. Ray, W.S. (1932). A preliminary study of fetal conditioning. Child Development, 3, 173–177. Ribeiro, A., Granier-Deferre, C., & Jacquet, A.-Y. (in preparation). Fetal discrimination of spectral information in speech. Richards, D.S., Frentzen, B., Gerhardt, K.J., McCann, M.E., & Abrams, R.M. (1992). Sound levels in the human uterus. Obstetrics & Gynecology, 80, 186–190. Richards, J.E. (1997). Localization of peripheral stimuli in infants: age, attention and individual differences in heart rate variability. Journal of Experimental Psychology: Human Perception and Performance, 23, 667–680. Richards, J.E., & Casey, B.J. (1991). Heart rate variability during attention phases in young infants. Psychophysiology, 28, 43–53.  2010 Blackwell Publishing Ltd.

Richards, J.E., & Casey, B.J. (1992). Development of sustained visual attention in the human infant. In B.A. Campbell, H. Hayne, & R. Richardson (Eds.), Attention and information processing in infants and adults (pp. 30–60). Hillsdale, NJ: Lawrence Erlbaum. Rosen, S., & Iverson, P. (2007). Constructing adequate nonspeech analogues: what is special about speech anyway? Developmental Science, 10, 165–169. Rotteveel, J.J., de Graaf, R., Stegeman, D.F., Colon, E.J., & Visco, Y.M. (1987a). The maturation of the central auditory conduction in preterm infants until three months post term. V. The auditory cortical response (ACR). Hearing Research, 27, 95–110. Rotteveel, J.J., Stegeman, D.F., & de Graaf, R. (1987b). The maturation of the central auditory conduction in preterm infants until three months post term. III. The middle latency auditory evoked response (MLR). Hearing Research, 27, 245–246. Rbsamen, R., & Lippe, W.R. (1998). The development of cochlear function. In E.W. Rubel, A.N. Popper, & R.R. Fay (Eds.), Development of the auditory system. Springer handbook of auditory research (pp. 193–271). New York: SpringerVerlag. Schleussner, E., & Schneider, U. (2004). Developmental changes of auditory-evoked fields in fetuses. Experimental Neurology, 190, 59–64. Scibetta, J.J., & Rosen, M.G. (1969). Response evoked by sound in the fetal guinea-pig. Obstetrics & Gynecology, 33, 830–836. Seebach, B.S., Intrator, N., Lieberman, P., & Cooper, L.N. (1994). A model of prenatal acquisition of speech parameters. Proceedings of the National Academy of Sciences, USA, 91, 7473–7476. Servire, J., Horner, K.C., & Granier-Deferre, C. (1986). Mise en vidence de l’activit fonctionnelle du systme auditif in utero du foetus de cobaye par autoradiographie au (14C) 2-dsoxyglucose. Comptes-rendus de l’Acadmie des Sciences, 302, srie III, 37–42. Shahidullah, S., & Hepper, P.G. (1994). Frequency discrimination by the fetus. Early Human Development, 36, 13–26. Shannon, R.V., Zeng, F.-G., Kamath, V., Wygonski, J., & Ekelid, M. (1995). Speech recognition with primarily temporal cues. Science, 270, 303–304. Smith, L.S., Dmochowski, P.A., Muir, D.W., & Kisilevsky, B.S. (2007). Estimated cardiac vagal tone predicts fetal responses to mother’s and stranger’s voices. Developmental Psychobiology, 49, 543–547. Smith, S.L., Gerhardt, K.J., Griffiths, S.K., Huang, X., & Abrams, R.M. (2003). Intelligibility of sentences recorded from the uterus of a pregnant ewe and from the fetal inner ear. Audiology & Neuro-Otology, 8, 347–353. Smith, Z.M., Delgutte, B., & Oxenham, A.W. (2002). Chimaeric sounds reveal dichotomies in auditory perception. Nature, 116, 87–90. Sohmer, H., Perez, R., Sichel, J.Y., Priner, R., & Freeman, S. (2001). The pathway enabling external sounds to reach and excite the fetal inner ear. Audiology & Neuro-Otology, 6, 109– 116. Sontag, L.W., & Wallace, R.F. (1935). The movement response of the human fetus to sound stimuli. Child Development, 6, 253–258. Sontag, L.W., & Wallace, R.F. (1936). Changes in the rate of the human fetal heart in response to vibratory stimuli. American Journal of Disabled Children, 51, 583–589.

352 Carolyn Granier-Deferre et al.

Spelt, D.K. (1938). Conditioned responses in the human fetus in utero. Psychological Bulletin, 35, 712–713. Starr, A., Amlie, R.N., Martin, W.H., & Sanders, S. (1977). Development of auditory function in newborn infants revealed by auditory brainstem potentials. Pediatrics, 60, 831–839. Vanhalewijn, M., & Moerman, J. (1971). Litla nornin Nanna. Idunn editor. van Woerden, E.E., van Geijn, H.P., Caron, F.J.M., van der Valk, A.W., Swartjes, J.M., & Arts, N.F. (1988a). Fetal mouth movements during behavioural states 1F and 2F. European Journal of Obstetrics & Gynecology and Reproductive Biology, 29, 97–105. van Woerden, E.E., van Geijn, H.P., Swartjes, J.M., Caron, F.J.M., Brons, J.T.J., & Arts, N.F. (1988b). Fetal heart rhythms during behavioral state 1F. European Journal of Obstetrics & Gynecology and Reproductive Biology, 28, 29–38. Vince, M.A., Billing, A.E., Baldwin, B.A., Toner, J.N., & Weller, C. (1985). Maternal vocalisations and other sounds in the fetal lamb’s sound environment. Early Human Development, 11, 164–170. Vouloumanos, A., & Werker, J.F. (2007). Listening to language at birth: evidence for a bias for speech in neonates. Developmental Science, 10, 159–171.

 2010 Blackwell Publishing Ltd.

Werner, L.A., & Gray, L. (1998). Behavioral studies of hearing development. In E.W. Rubel, A.N. Popper, & R.R. Fay (Eds.), Development of the auditory system. Springer Handbook of auditory research (pp. 12–80). New York: SpringerVerlag. Woods, J.R., & Plessinger, M.A. (1989). Fetal sensory sequencing: application of evoked potentials in perinatal physiology. Seminars in Perinatology, 13, 380–392. Woodward, S.C., & Guidozzi, F. (1992). Intrauterine rhythm and blues? British Journal of Obstetrics & Gynaecology, 99, 787–790. Zatorre, R.J., Belin, P., & Penhune, V. (2002). Structure and function of auditory cortex: music and speech. Trends in Cognitive Sciences, 6, 37–46. Zimmer, E.Z., Fifer, W.P., Kim, Y-I., Rey, H.R., Chao, C.R., & Myers, M.M. (1993). Response of the premature fetus to stimulation by speech sounds. Early Human Development, 33, 207–215. Received: 31 October 2008 Accepted: 8 March 2010