Our knowledge of infant perception and cognition is primarily based on habituation and dishabituation, but the underlying neural mechanisms for these processes per se remain unclear. It has been argued that habituation is related to building internal representations of repeated stimuli in the central nervous system, whereas dishabituation is related to an increased attention to novel items and events. This leads to a hypothesis that a distributed network including the sensory, association and prefrontal cortical regions of young infants is involved in those processes, in contrast with the classical developmental view that onset of the functioning of the prefrontal cortex is delayed. Here we examined the time evolution of spatio-temporal hemodynamic responses related to the auditory habituation and dishabituation in the temporal and prefrontal regions of 3-month-old infants by using multichannel near-infrared spectroscopy. We found that the temporal regions remained activated by repetitive auditory stimuli; however, the prefrontal regions exhibited phasic activation in relation to novel stimuli. The dissociated activation pattern between the temporal and prefrontal regions suggests that distinct cortical regions play distinct functional roles in auditory habituation and dishabituation, and that the prefrontal cortex is involved in perceiving invariance or novelty of the immediate environment in early infancy.
Habituation and dishabituation, decreased response to repeated stimuli and recovery of response to novel stimuli, respectively, are seen for a wide variety of organisms ranging from Aplysia to human (Fantz 1963; Thompson and Spencer 1966; Pinsker et al. 1970). As the habituation effect is sustained from minutes to days, habituation is thought to be the elementary form of learning and memory (Thompson and Spencer 1966; Pinsker et al. 1970). In particular, it has been argued that habituation is not merely sensory fatigue but is related to building internal representations of repeated stimuli in the central nervous system, whereas dishabituation is related to an increased attention to novel items and events (Sokolov 1963). On the basis of such assumptions, these behavioral responses are widely used as an experimental paradigm in infant studies investigating perception, cognition, and language development (Fantz 1963; Kellman and Spelke 1983; Mehler et al. 1988). Despite that much attention has been given to young infants’ behavior and brain development, the underlying neural mechanisms of infant habituation and dishabituation remain unclear.
The functional magnetic resonance imaging (fMRI) studies of adult subjects have shown that although the sensory and association cortices are involved in the processing of repeated stimuli (Grill-Spector et al. 1999; Dehaene-Lambertz et al. 2006), the prefrontal cortex and hippocampal regions are involved in the detection of novel stimuli and events (Doeller et al. 2003; Yamaguchi et al. 2004). Thus, we hypothesize that a distributed network including the sensory, association and prefrontal cortical regions is involved in habituation and dishabituation in 3-month-old infants. It conflicts with the classical developmental view that onset of the functioning of the prefrontal cortex is delayed due to the prolonged development involving myelination and synaptogenesis in the prefrontal cortex, as compared with the primary sensory cortices (Flechsig 1901; Huttenlocher 1979; Huttenlocher and Dabholkar 1997). However, behavioral studies show that young infants exhibit sophisticated actions that may require cortical mechanisms for attention and working memory (Gilmore and Johnson 1995; Rosander and von Hofsten 2004). Moreover, an event-related potentials study with 3-month-old infants showed a frontal response to novel stimuli, suggesting the existence of an anterior neural network in the brain (Dehaene-Lambertz and Dehaene 1994). Thus, we expected that the higher cortical regions are involved in infant habituation and dishabituation.
Although much attention has been directed to the functional development of the cortex (Johnson 2001), limited information is available due to the limitations of brain imaging methods when used on young infants. In the present study, we used the neuroimaging technique of near-infrared spectroscopy (NIRS) (Obrig and Villringer 1997; Villringer and Chance 1997). This technique permits noninvasive and safe measurement of cerebral blood oxygenation associating with cortical activations in young infants (Meek et al. 1998; Peña et al. 2003; Taga et al. 2003; Kotilahti et al. 2005; Homae et al. 2006; Bortfeld et al. 2007; Minagawa-Kawai et al. 2007; Taga and Asakawa 2007). Thus, it is possible to examine if separate regions of the infant cortex show distinct time evolutions of activations in related to habituation and dishabituation across consecutive presentations of auditory stimuli over several minutes.
Materials and Methods
Eighty full-term healthy Japanese infants ranging in age from 3 to 4 months (boys: 48; girls: 32; mean age 115.4 days; range 92–142 days; SD 10.8) participated in the present study. Forty infants were allocated to the experimental group (boys: 24; girls: 16; mean age 114.4 days; SD 12.3) and the other forty to the control group (boys: 24; girls: 16; mean age 116.7 days; SD 9.0). An additional 63 infants were tested but excluded from the analysis because of awaking during the experiment (n = 34), head movements producing large motion artifacts in the signals (n = 25), and failure in probe placement due to obstruction by hair (n = 4). Informed consent was obtained from the parents of infants before the initiation of experiments. These experiments were approved by the ethical committee of the Graduate School of Education (University of Tokyo).
On each trial, 10 identical syllables were presented for 5 s followed by 15 s of silence. We prepared 2 syllables (/ba/ and /pa/) as the auditory stimuli; one for the habituation stimuli and the other for the novel stimuli, counterbalanced across infants. These syllables have similar frequency bands but the consonant [b] can be distinguished from [p] based on a temporal cue called voice onset time defined as the time between the release of a consonant and the onset of vocal fold vibration. A previous study using the habituation and dishabituation paradigm revealed that 1- and 4-month-old infants can discriminate these 2 syllables (Eimas et al. 1971). We recorded the syllables as naturally produced by a female Japanese speaker (44 100 Hz, 16 bit). The duration of both syllables was 250 ms. Stimuli were presented via a loudspeaker system (BOSE MMS-1) with a maximum amplitude of 56 dB.
We designed a procedure to reveal the cortical bases of auditory habituation and dishabituation as shown in Figure 1A. All infants were initially exposed to the same stimuli for 15 trials in the habituation phase (Hab1, Hab2, and Hab3). After that, half of the infants (experimental group) were successively presented with novel stimuli for 5 trials in the test phase (Test) and then exposed to the same stimuli as were presented in the habituation phase again for 5 trials in the post-Test phase (Post-test). The other half of the infants (control group) was exposed to the identical stimuli throughout all phases. Total measurement time was 8.5 min.
Infants were held by an experimenter who sat on a chair in front of the speaker in a sound attenuated room. We tested in daytime although infants were sleeping soundly and were almost motionless. The infants’ sleeping state corresponded to quiet sleep (QS) of infants. QS is the dominant and stable state in daytime sleep of young infants around 3 months (Peirano et al. 2003). Recent infant studies using NIRS have demonstrated cortical activations of young infants in response to auditory stimuli while sleeping (Pena et al. 2003; Kotilahti et al. 2005; Homae et al. 2006). Furthermore, sleeping infants show responses related to the detection of changes in stimulus (Cheour et al. 2000). Thus, we expected that sleeping infants would show cortical responses to auditory sounds. The behavior of infants was video-recorded to identify the trials in which infants slept without moving their head.
We used 2 NIRS instruments (ETG-100, Hitachi Medical Corporation, Tokyo, Japan), one for each hemisphere. These instruments generated 2 wavelengths of near-infrared light (780 and 830 nm) through optical fibers and measured the time courses of changes in oxyhemoglobin (oxy-Hb) and in deoxyhemoglobin (deoxy-Hb) at 48 channels with 0.1 s time resolution. Because the precise optical path length was unknown, the unit of these values was a molar concentration multiplied by length (mM·mm). We placed 4 sets of 3 × 3 arrays with 5 incident and 4 detection fibers mounted on a flexible cap over the temporal and prefrontal regions of the right and left hemispheres (Fig. 1B). The distance between adjacent incident and detection positions was set at 2 cm (Taga et al. 2007), and therefore a 4 × 4 cm area for each of the temporal and prefrontal regions was covered. The intensity of the illumination was set at 0.6 mW. The measured area in each hemisphere was correctly positioned by using skull landmarks of the nasion, vertex and external auditory canals in each infant. Measurement channel 4 (Ch4) and Ch9 on the right hemisphere, and Ch40 and Ch45 on the left hemisphere were set on a coronal line from the vertex to the external auditory pores. The right Ch9 and left Ch45 were set 1 cm above the T4 and T3 positions, respectively, of the international 10–20 system of electrode placement. The right Ch24 and left Ch35 were set on the Fp2 and Fp1 positions, respectively. Studies in adults show that the T3 and T4 positions are projected around the middle and superior temporal gyri (Okamoto et al. 2004). A recent NIRS study with young infants also shows that right Ch9 and left Ch45 are significantly activated in response to auditory stimuli (Taga and Asakawa 2007).
Data Processing and Analysis
We rejected trials that included large motion artifacts due to the head movements as identified by monitoring the video recordings and analyzing rapid changes in time series of oxy-Hb signals. We excluded infant data that contained any one epoch with less than 3 good trial data. Following these screening process, we obtained 80 infants’ data for further analysis. The mean numbers of trials for each epoch (Hab1∼3, Test, and Post-test) were 4.8, 4.6, 4.7, 4.5, and 4.6 in the experimental group and 4.4, 4.3, 4.4, 4.4, and 4.3 in the control group, respectively. Then, we also eliminated any channel with a low signal- to-noise ratio due to obstruction by hair. The mean numbers of infants per channel were 38.8 (range 31–40) for the experimental group and 38.8 (range 33–40) for the control group.
Next, we corrected the baseline using linear fitting to a mean signal of the 10 time points (1 s) before stimulus onset and that of the 10 time points at 19 s after onset in each trial data. In further analysis, we focused on oxy-Hb signals, which estimate the regional cerebral blood oxygenation changes during brain activation because of their superior signal-to-noise ratio for oxy-Hb to deoxy-Hb (Strangman et al. 2002). In each of the individual infant data, by extracting each trial data from the time series and averaging them under each epoch, we obtained the hemodynamic responses for each channel. Then, we determined an averaged time course among all trials of all measurement channels across both groups and calculated the time to peak in the time course. We analyzed the time points that were larger than one standard deviation among the averaged time course, and calculated mean signal changes in this time window for each channel of individual subjects in the statistical analyses that followed.
To identify the activated regions under each epoch, we considered the individual data as random effects and performed t-tests for each channel. First, a t-test of the signal changes under each epoch was performed against zero baselines. Then, we analyzed the cortical regions showing the decreased magnitude of responses by stimulus repetition in the habituation phase. Two-way ANOVAs were conducted with epoch (Hab1, Hab2, and Hab3) and group (experimental and control) as factors and we used the Tukey-Kramer test for post hoc analysis. Next, we identified the locus of dishabituation by comparing mean signal differences between Test and Hab3 in the experimental group with those in the control group on a channel-by-channel basis (2 samples t-test). The same comparisons between groups were done with the signal differences between Post-test and Hab3. To resolve statistical problems of multiple comparisons for multichannel NIRS, we applied the false discovery rate (FDR) correction (Benjamini and Hochberg 1995; Singh and Dan 2006). In the present study, we used the median adaptive FDR correction with α = 0.05 (Benjamini et al. 2006). Although the FDR correction allowed us to distinctively set a statistical threshold for each analysis, we applied a single criterion (P < 0.01), the most conservative threshold, in all analyses.
Figure 2 shows the entire time series of relative changes of oxy-Hb and deoxy-Hb averaged among infants for each group in the temporal (Ch9) and the prefrontal (Ch15) channels. The temporal channel showed a distinctive increase of oxy-Hb signals and a slight decrease of deoxy-Hb signals in response to auditory stimuli for every trial. In contrast, the hemodynamic response of the prefrontal channel rapidly diminished after first few trials and resumed only in Test of the experimental group. In further analysis, we focused on oxy-Hb signals because of their superior signal-to-noise ratio for oxy-Hb to deoxy-Hb.
By averaging the oxy-Hb signals over data blocks under each epoch in the experimental and control groups, we found a spatio-temporal change of hemodynamic responses related to the auditory habituation and dishabituation (Fig. 3). In both groups, the bilateral temporal regions showed large signal changes in all epochs, but the intensity of signal changes were different between epochs. The prefrontal regions also exhibited positive changes of oxy-Hb signals in Hab1 of both groups, but such obvious changes were not found in both Hab2 and Hab3. In Test, the novel stimuli caused different hemodynamic changes between groups; the prefrontal regions of the experimental group exhibited positive oxy-Hb changes, whereas the same regions of the control group did not show any changes.
To quantify the event-related hemodynamic response to the auditory stimuli, we analyzed mean time courses of hemodynamic response among all trials of all measurement channels across both groups. The maximum increase was reached at 7.4 s with a signal change of 0.0146 mM·mm (Fig. 4). The amount of signal changes between 4.8 and 9.6 s in the averaged time course was larger than one standard deviation among the time points, and we used this time window to calculate mean signal changes for each channel of individual subjects.
Next, to examine changes in activated cortical regions associated with habituation and dishabituation to auditory stimuli, we performed t-test (one tail) on the mean changes in oxy-Hb signals against zero baseline under each epoch. The multiple comparisons among the 48 channels were corrected by adopting the median adaptive FDR correction with α = 0.05 in each t-test, and we applied the single criterion of the most conservative threshold of P < 0.01 in all t-tests. Dissociated activation patterns were found between the temporal and prefrontal regions during auditory habituation and dishabituation in young infants (Fig. 5). In the experimental group, the auditory stimuli significantly activated broad regions over the bilateral temporal and prefrontal cortices in Hab1. In Hab2 and Hab3, the repetitive stimuli significantly activated only the local temporal regions of both hemispheres. The measurement channels Ch6 and Ch9 in the right hemisphere and Ch 43 and Ch45 in the left hemisphere showed greatest activations in Hab1 [Ch6: t(38) = 6.82, P < 10−7; Ch9: t(38) = 8.25, P < 10−9; Ch43: t(35) = 5.36, P < 10−5; Ch45: t(30) = 5.51, P < 10−5] and remained significantly activated in Hab2 [Ch6: t(38) = 3.40, P < 10−3; Ch9: t(38) = 3.31, P < 10−2; Ch43: t(35) = 3.70, P < 10−3; Ch45: t(30) = 4.34, P < 10−4] and Hab3 [Ch6: t(38) = 5.31, P < 10−5; Ch9: t(38) = 4.36, P < 10−4; Ch43: t(35) = 4.81, P < 10−4; Ch45: t(30) = 3.66, P < 10−3]. These measurement channels lay in a symmetrical line over both hemispheres and previous NIRS studies with young infants also show that these channels are significantly activated in response to auditory stimuli (Homae et al. 2006; Taga and Asakawa 2007). Thus, the bilateral temporal regions that remained activated by repetitive auditory stimuli were thought to be the auditory area. The similar spatio-temporal changes of cortical activations were also observed in the control group. In Test, the novel stimuli reactivated broad regions over the bilateral temporal and prefrontal cortices of the experimental group. In the temporal regions, the measurement channels in the auditory areas, especially for the right hemisphere, showed highly significant activations [Ch6: t(38) = 6.11, P < 10−6; Ch9: t(38) = 7.07, P < 10−8; Ch43: t(35) = 4.62, P < 10−4; Ch45: t(30) = 4.04, P < 10−3]. In the prefrontal regions, the measurement channels of the right hemisphere Ch15 [t(39) = 3.67, P < 10−3), Ch20 [t(39) = 3.39, P < 10−3], and Ch18 [t(39) = 3.52, P < 10−3] showed the greatest activations. In Post-test, the stimuli, which were changed back to the familiarized stimuli, caused no significant activations in the prefrontal regions. In contrast to the experimental group, only local activations of the temporal regions were observed in Test of the control group.
To identify the cortical regions showing the decreased magnitude of responses by stimulus repetition in the habituation phase, we conducted 2-way ANOVAs with epoch (Hab1, Hab2, and Hab3) and group (experimental and control) as factors (P < 0.01). In the habituation phase, we observed a significant main effect of epoch in the cortical responses for most of the measurement channels, but no differences were found between groups (Fig. 6). These channels showed a significant decrease in the magnitude of responses from Hab1 to Hab2 and/or to Hab3 by post hoc analysis using the Tukey-Kramer test (P < 0.05). But the measurement channels Ch43 and Ch45, which were located in the auditory areas of the left hemisphere and remained significantly activated in Hab2 and Hab3, did not show a significant decrease of the magnitude of responses.
In order to identify the locus of dishabituation, the mean signal differences between Test and Hab3 were directly compared between the experimental and control groups for each channel (2 samples t-test, one tail, P < 0.01). Figure 7 shows that the bilateral prefrontal regions showed highly significant activation to the novel stimuli [Ch33: t(78) = 3.02, P < 0.001; Ch15: t(78) = 2.79, P < 0.002; Ch16: t(78) = 2.52, P < 0.004; Ch30: t(78) = 2.49, P < 0.004]. By contrast, using the same statistical analysis for the signal differences between Post-test and Hab3, no significant difference was observed between the experimental and control groups. Although a slight increase in number of activated channels against zero baseline was observed in Post-test compared with Hab3 and Test in the control group, no significant difference between the groups suggests that the change in number of activated channels was within time fluctuation of activation to the same stimuli. This further suggests that the habituation effect was lasting even after the intervening 5 trials involving the novel stimuli (about 100 s).
Figure 8A shows the time evolutions of the waveform of the hemodynamic responses in the selected measurement channels that showed greatest activation in Hab1 and Test at the temporal (Ch9) and prefrontal regions (Ch15). Conspicuous differences were found between the temporal and prefrontal channels (Fig. 8A). The temporal channel showed prominent responses to the auditory stimuli across all phases in both groups. On the other hand, the prefrontal channel showed distinct responses under Hab1 in both groups and Test in the experimental group. By plotting the evolution of the intensity of the responses, we obtained habituation and dishabituation curves (Fig. 8B). In the experimental group, the activities in the temporal channel were sustained across all epochs with decreasing magnitude of response in Hab2 and increasing in Test. By contrast, the responses in the prefrontal channel dropped down to around zero after Hab1 and resumed only in Test. Such resumed activations in Test were not observed in the control group.
A crucial aspect of our findings was the dissociated activation pattern in the prefrontal region; activation in response to novel stimuli and silence in the fully habituated state. This indicated that the prefrontal region is involved in perceiving invariance and novelty in the immediate environment. The temporal regions, by contrast, were maintained in a state to be activated by auditory stimuli regardless of the number of repetitions and the feature of the stimuli. It should be noted that the amplitudes of the responses of the temporal regions were modulated in accordance with habituation and dishabituation. A possible explanation for this modulation of response in the temporal regions is that the auditory areas have intrinsic mechanisms for changes in activation associated with encoding and discrimination of auditory inputs. Another possible explanation is that the prefrontal regions provide top-down modulation of the auditory processing in the temporal regions. In either case, the results of the present study suggested that distinct cortical regions exhibit distinct functional specificity very early in life. In adults, novel stimuli activate a distributed network including the prefrontal cortex, which is the key element orienting responses to stimuli and is thought to be associated with enhancing the sensory processing and memorizing of novel stimuli (Kirchhoff et al. 2000; Doeller et al. 2003; Yamaguchi et al. 2004). Considering these adult studies, we postulated that an inter-regional network including the temporal and prefrontal regions underlies the auditory habituation and dishabituation in infants. This interpretation finds support in a recent study using diffusion tensor imaging and tractography that demonstrated that the white matter fiber bundles connecting the temporal and prefrontal cortices are already in place in early infancy despite immature myelination (Dubois et al. 2006).
For several decades, the habituation paradigm has been an important behavioral tool for assessing the perceptual and cognitive ability in infants. In this study, we revealed that the cortical bases of habituation and dishabituation to auditory stimuli over a time scale similar to that of the behavioral studies in infants aged 3 months. As a stimulus became familiar, the global activation pattern over the temporal and prefrontal regions changed to the local activation pattern of the temporal regions. The decrease in the cortical activations was stimulus-selective, because the similar but the novel syllable increased cortical activations over the global regions again. This suggested that infant habituation occurs as a consequence of constructing the stimulus-selective representation. A number of previous studies in animals and human adults have also proposed that suppression of neural responses to repeated stimuli can be associated with the formation of selective neural representations (Desimone 1996; Wiggs and Martin 1998). Furthermore, the prolonged suppression effect over the post-test phase in the present study demonstrated that neural representation of stimuli is maintained for a few minutes after the habituation phase. As well, considering the behavioral results that the recognition memory of young infants is retained for at least a few days (Fagan 1973; Martin 1975), habituation may closely relate to the process of forming memory in the early months of life.
In the present work, the prefrontal activations were observed in sleeping infants. A fMRI study in adults reported that, whereas the prefrontal cortex was less activated during deep sleep compared with wakefulness, the salient auditory stimuli produced increased activation in the dorsolateral prefrontal cortex during sleep (Portas et al. 2000), suggesting that the processes of attentional operations remain during sleep. In addition, mismatch negativity elicited by novel auditory stimuli in wakefulness is also present during sleep in young infants (Cheour et al. 2000). Hence, we can assume that the functioning of the infant brain is not totally different between the sleep and awake states.
The results that we present here for the activation of the prefrontal regions in 3 month olds require that there is a change in how cortical development in early infancy is viewed. Whether the functional brain development proceeds hierarchically in accordance with the anatomical brain development (Flechsig 1901; Huttenlocher 1979; Huttenlocher and Dabholkar 1997) or concurrently via inter-regional interactions over whole cortical areas (Rakic et al. 1986; Rakic et al. 1994; Johnson 2001) has been a topic of debate. Although the process of cortical development before 3 months of age has not been revealed, the present study infers that a functional network including the prefrontal region may concurrently evolve through inter-regional interactions early in life.
We thank ES Spelke for comments on an earlier version of this manuscript, H Koizumi, A Maki, and Y Konishi for helpful discussions, N Ichikawa, M Fujiwara, and T Ishizuka for technical support of the NIRS instrument, and K Asakawa for technical assistance. Conflict of Interest: None declared.