This work was supported by U.K. Medical Research Council Component Grant G0400120, Programme Grant 9715587, and a studentship. We would like to thank Gergely Csibra, Fani Deligiani, and Jem Hebden for comments and input on the methods and earlier versions of the draft. We would also like to thank all of the parents and their infants who participated in the study.
concerning this article should be addressed to Sarah Lloyd-Fox, Centre for Brain and Cognitive Development, Birkbeck, University of London, London, United Kingdom. Electronic mail may be sent to email@example.com.
The capacity to engage and communicate in a social world is one of the defining characteristics of the human species. While the network of regions that compose the social brain have been the subject of extensive research in adults, there are limited techniques available for monitoring young infants. This study used near infrared spectroscopy to investigate functional activation in the social brain network of 36 five-month-old infants. We measured the hemodynamic responses to visually presented stimuli in the temporal lobes. A significant increase in oxyhemoglobin was localized to 2 posterior temporal sites bilaterally, indicating that these areas are involved in the social brain network in young infants.
While the network of regions that together compose the social brain have been the subject of extensive research in adults, their origins in infancy largely remain obscure. Part of the reason for this lack of knowledge is methodological. For a variety of reasons, it is very difficult to scan awake and conscious infants in functional magnetic resonance imaging (fMRI) protocols that involve visual presentations. Further, while electroencephalography (EEG) measures give good temporal resolution, their spatial resolution remains comparatively poor. A recently developed noninvasive imaging method, near infrared spectroscopy (NIRS), can potentially bridge this methodological gap. In this study, we demonstrate that NIRS can be used to investigate visually induced activity in regions of the social brain in young infants.
Within the extent of the STS, different areas appear to be specialized for different types of stimuli. Studies that involve perceived eye movements find a focused area of bilateral activation around the posterior STS, often in the right hemisphere, while responses to hand and mouth movements show more widespread activation extending anteriorly, and often in the left hemisphere (for a review of several studies, see Allison, Puce, & McCarthy, 2000; Pelphrey, Morris, Michelich, Allison, & McCarthy, 2005). Moreover, research suggests that the right posterior STS may be more sensitive than the left to the type of social motion perceived—mutual versus averted gaze (Pelphrey, Viola, & McCarthy, 2004), congruent versus incongruent gaze (Mosconi et al., 2005), and expressive versus instrumental gestures (Gallagher & Frith, 2004).
While much fMRI research has been performed on biological motion processing and associated activity in the temporal posterior lobe in adults, there are limited techniques available for monitoring functional brain activation in infants. Specific spatial localization of brain activity has not been investigated extensively during infancy because techniques such as fMRI cannot accommodate mobile, awake infants. NIRS is a noninvasive optical technique that provides the opportunity to measure the hemodynamic response to neuronal activation in a range of different subjects including awake infants. With this optical technique, the light migrates from sources on a sensor pad located on the head, through the skin, skull, and underlying brain tissue, and is then detected by sensitive detectors on the same sensor pad (for early NIRS work, see Chance, Zhuang, UnAh, Alter, & Lipton, 1993; Jöbsis, 1977). Changes in blood oxyhemoglobin (HbO2), deoxyhemoglobin (HHb), and total hemoglobin (HbT = HbO2 + HHb) in the underlying cortex are measured by detecting changes in reflected near infrared light. The concentration of these chromophores (the molecules that are responsible for the color of the blood due to their absorption of light at different wavelengths) change according to the metabolic demand of the neurons in a given cortical region, thus enabling functional brain activation to be measured. A typical hemodynamic response to cortical activation is an increase in blood flow leading to an increase in HbO2 and a (relatively smaller) decrease in HHb as it is displaced from the veins, leading to an increase in HbT. Previous work suggests that the properties of the vascular response measured by NIRS is comparable to the blood oxygen level dependent (BOLD) response seen in MRI research. However, there is ongoing debate concerning which chromophore changes most closely map the BOLD response (for review, see Ferrari, Mottola, & Quaresima, 2004; Obrig & Villringer, 2003). Although there is a causal relation between changes in HHb concentration and the BOLD signal, Strangman, Culver, Thompson, and Boas (2002) found the strongest correlation between BOLD fMRI changes and HbO2. Moreover, while adult NIRS studies show HbO2 and HHb responses consistent with BOLD fMRI signal changes, infant data reveal a less consistent pattern of activation (Baird et al., 2002; Meek et al., 1998; Sakatani, Chen, Lichty, Zuo, & Wang, 1999; Wilcox, Bortfeld, Woods, Wriuck, & Boas, 2005; Zaramella et al., 2001). Therefore, in NIRS research, it is appropriate to investigate both HbO2 and HHb changes in response to a given stimulus.
Initial studies using single-channel NIRS to monitor functional activation in infants used standard visual (checkerboard) and auditory (pure tone) stimuli (Meek et al., 1998; Sakatani et al., 1999). More complex stimuli have now been investigated using a range of different multichannel NIRS optical topography systems (i.e., Blasi et al., 2007; Otsuka et al., 2007; Peña et al., 2003; Shimada & Hiraki, 2006). Whereas single-channel systems were limited to measuring just one location, multichannel optical topography allows simultaneous measurements across an array of channels, enabling spatially localized activation across a larger area of the cortex to be investigated. In addition, by adjusting the configuration of the sources and detectors it is possible to potentially discriminate responses at different depths of the cortex by using channels with several different separations. For example, in a recent study on face perception, we found a more widespread hemodynamic response (increase in HbO2) to faces compared to visual noise in the occipital area in 4-month-old infants (Blasi et al., 2007). Furthermore, a larger number of channels with a significant increase in HbO2 were found at the deepest depth, suggesting that there may be some discrimination in the response as a function of depth relative to the cortical surface.
Multichannel arrays may also be used in several locations simultaneously to measure differential responses over differing brain regions. A recent study used two arrays located over the temporal lobes to investigate neural activation in 5- to 8-month-old infants. Otsuka et al. (2007) found differential responses to face inversion between hemispheres in response to static images of upright and inverted faces. For upright faces an increase in HbT and HbO2 was observed in the right hemisphere indicating a lateralized inversion effect. Indeed, the authors suggest that this response could reflect activity associated with face perception in the STS. Although this study investigated static faces without implied motion, the rapidly changing face stimuli within trials (five faces per 5 s trial) could have caused a response in the STS. However, limited conclusions can be drawn about the spatial localization of the response as the reported effects are based on grouped data across all channels in each lateral array. The findings support the need to further investigate the temporal lobes, and the role it plays in an infant’s social brain network.
The purpose of the current study was to investigate whether young infants show a hemodynamic response over the posterior temporal lobe to a complex social stimulus involving biological motion. To maximize the potential response and maintain the infants’ attention, the biological motion condition composed of naturalistic video clips of human movements involving eyes, mouth, and hands. In Experiment 1, this was compared with complex but static images of transport, which served as the baseline condition. In Experiment 2, a dynamic nonsocial stimulus was introduced. We chose to study 5-month-old infants to facilitate comparison with previous NIRS studies. To investigate this question, two sensor pads were placed over the temporal lobe on each hemisphere. The posterior half of the sensor pad lay over the scalp locations T3–T5/T4–T6 while the anterior half of the pad extended toward F7/F8 (according to the 10–20 system; Jasper, 1958). The placement of the sensor pads allowed us to make predictions about the spatial localization of the response using the findings of adult fMRI studies in this field. If the infants’ functional activation to biological motion perception resembles that of adults’, then we hypothesize that localized activation will occur in response to the social stimuli in the posterior half of each sensor pad near to scalp locations T5 and T6. As the stimuli used in this study included eye, mouth, and hand movements, we further predicted that there will be bilateral activation rather than any overall lateralization effect as has been suggested in earlier research (for review, see Allison et al., 2000; Pelphrey et al., 2005). However, if 5-month-old infants do not yet possess a specialized STS, we predict no differences in activation between the human moving video clips and the control static transport images.
Participants. Twenty-four 5-month-old infants took part in the study (13 females and 11 males; age, 154 ± 8.7 days old). Another 2 infants participated but were excluded from the study as they failed to look at the stimuli for the minimum number of six trials; all other infants completed the whole study. All parents volunteered by responding to advertisements and gave written, informed consent to participate. The study protocol was approved by the appropriate Ethics Committee. The infants were from a varied ethnic and socioeconomic background, predominantly White (British/non-British) or British Black (mixed/African/Caribbean).
Procedure. The infants sat on their parent’s lap and were encouraged to watch the stimuli displayed on a 46-in. plasma screen with a viewing distance of approximately 100 cm. The experiment ended when the infants became bored or fussy. The experimental condition consisted of full-color, life-size social video clips of female actors who either moved their eyes left or right, their mouth in silent vowel movements, or performed hand games—“peek-a-boo” and “incy wincy spider.” These were presented in an ABA format (6, 4, and 6 s, designed to keep the infant’s attention). The baseline condition consisted of full-color, still images of different types of transport (i.e., cars and helicopters) presented randomly for a pseudorandom duration (1–3 s). These images were selected to be colorful, complex, and interesting, and ensured that infants remained attentive to the screen. Note that NIRS studies with adults use a blank screen as the baseline but this is not possible when working with infants; therefore, the static images act as the baseline for the activated period containing the social video clips. The overall surface area of the displayed experimental stimuli and baseline stimuli were equivalent and subtended a visual angle of approximately 12°.
The sequence of stimulus presentation is illustrated in Figure 1. The session began with a rest period (30 s minimum) during which the infant was shown shapes in the four corners of the screen to familiarize them with the general experimental setup. Following this, the trials alternated one after the other, that is, a 16-s experimental trial followed by a 16-s baseline trial. In addition, during every third trial, music was played to help maintain the infant’s overall engagement with the task. Thus, music was played at an approximately equal number of experimental and baseline trials for each infant. If necessary, occasional alerting sounds were used to draw the infant’s attention back to the screen. The experimental trials and baseline trials were presented consecutively three times each, followed by a preferential looking trial. For this, a social still image and a transport still image of equal size were presented for 5 s, one on the left and one on the right of the screen. Following this, a further six consecutive experimental and baseline trials were presented, and then a further preferential looking trial (this time with the human and transport images displayed on the opposite side of the screen to the first preferential looking trial). Thereafter, consecutive experimental and baseline trials were displayed repeatedly until the infant became bored or fussy (see Figure 1 for an overview of this procedure).
Data acquisition and array placement. To investigate cortical activation, NIRS measurements were made using the University College London (UCL) topography system (Everdell et al., 2005). The multichannel system uses two wavelengths at 770 and 850 nm in a frequency multiplexed approach allowing rapid data acquisition of the attenuation signal from the reflected near infrared light. The array of channels are designed and adapted for each study protocol, allowing flexibility in the source--detector geometry and locations of the arrays. Eight optodes, in a 10-channel (source--detector pairs) arrangement with an inter-optode separation of 20 mm, were placed on the temporal lobe on each hemisphere in custom-built arrays and headgear (see Figure 2). The midpoint of the lower row of channels lay over scalp location T3 on the left hemisphere and T4 on the right. The schematic map (Figure 2) shows an estimation of the 10–20 system scalp locations, using available anatomical landmarks for the 5-month-old infants. The posterior area of each pad lies approximately over the scalp locations T5/T6, analogous to the region of interest (the STS area reported in adult MRI studies).
Data rejection. To assess whether the infants were looking during each trial, we recorded their eye movements and coded looking time off-line. This was the first step in the selection of valid trials. For a trial to be considered valid, the infant had to be looking at the screen for at least 4 s prior to the experimental trial onset and for a minimum of 80% of the following experimental trial period. A minimum of six valid experimental trials were required to include an infant in the study.
Following this, the recorded NIR attenuation measurements for each infant were analyzed and trials or channels were rejected from further analysis based on the quality of the signals. Criteria for channel rejection included the presence of: (a) large movement artifacts assessed by measuring the coefficient of variation (CV) of the signal. Channels were excluded if the CV of the attenuation measurement for each wavelength exceeded 10% or if the difference in CV between the attenuation measurements for the two wavelengths (|CV770–CV850|) exceeded 5%. These changes in CV could be due to movement of the pads and hat, differential occlusion of the source fibers for each wavelength or a loose fiber in the pad, or (b) high-frequency noise beyond the limits of physiological effects, where the normalized high frequency power is greater than 35% of the total power of the signal (see Kirjavanien et al., 2001). For each infant, the channels that survived these rejection criteria were analyzed for trial selection. The trial selection analysis identified sharp changes in the signal caused by sudden movements rather than a generally noisy channel from a continuously moving infant as identified by measuring the coefficient of variance. Following data conversion from attenuation to concentration data (a full description is given in the following Data Analysis section), trials that contained changes in HbO2 concentration that exceeded a predefined range (±5 μM during the baseline trials and ±15 μM during the experimental trials where artifacts in the signal may occur in addition to activation), were removed from the data set. The minimum number of valid experimental trials for each channel was six.
Data analysis. For each infant, the attenuation signal (from the reflected near infrared light) was low-pass filtered, using a cutoff frequency of 1.8 Hz. The data were then divided into blocks consisting of 4 s of the baseline trial (still transport images) prior to the onset of the 16 s experimental trial (human video clips), plus the following 16 s baseline trial (still transport images). This 36-s block of attenuation data were detrended with a linear fit between the first and last 4 s of the 36 s block. Within these 4 s segments, we assume that all effects of the experimental stimulus have subsided (Blasi et al., 2007). The data were then converted into changes in concentration in HbO2 and HHb using the modified Beer–Lambert law (Delpy et al., 1988) and assuming a differential pathlength factor for infants (5.13; Duncan et al., 1995).
Following this, valid experimental stimulus trials were averaged together for each infant, and a time course of the mean change in HbO2 and HHb was compiled for each channel. These average time courses for each infant were then compiled into a grand averaged time–response curve of the hemodynamic response (across all infants) for each channel. A time window (region of interest) was selected between 10 and 18 s postexperimental stimulus onset. This period of time was selected to include the range of maximum changes observed across infants for HbO2 and HHb. Statistical comparisons of the response to experimental versus baseline trials across all infants were made using the valid data for each channel. A one-sample t test was performed during the specified experimental trial time window to compare the maximum change (or amplitude) in HbO2 and HHb with the mean of the baseline trial.
According to the criteria previously described, valid data were obtained from 24 of the 26 infants tested. The mean number of experimental trials recorded in these infants was 9.6 and the mean length of the recording session was 6 min 30 s. The proportion of valid channels across the group of infants was 76% in the left lateral pad and 82% in the right lateral pad. The preferential looking task indicated that there was a significant preference for the experimental stimulus over the baseline stimulus (exp = 80.2%, baseline = 19.8%, p < .001, t = 5.766, paired t test).
A one-sample t test of the maximum increase (or amplitude) in HbO2 in response to the experimental stimuli (during the specified time window 10–18 s poststimulus onset) revealed a significant increase from the baseline in HbO2 in two channels (Channel 14: t = 2.356, p < .05, df = 18; Channel 28: t = 3.117, p < .01, df = 23; note that the degrees of freedom differ as recordings on Channel 14 were achieved in only 19 of the total 24 infants). These two channels were located in the most posterior area of the pads, over the T5 and T6 scalp locations (see Figure 1). Within these two channels, a sign test revealed that across the group, infants were significantly more likely to have an increase in HbO2 in response to the experimental stimuli (Channel 14: n = 18, p = .048; Channel 28: n = 24, p = .01) There were also an additional two channels that reached borderline significance at the most anterior area of the pads (Channel 2: t = 2.037, p = .055, df = 16; Channel 16: t = 1.83, p = .08, df = 20). There were no significant changes in the concentration of HHb across infants. Table 1 gives a summary of these results.
Table 1. Mean Maximum Changes in HbO2 and HHb Concentration (μΜ) Across Infants During the Window of Activation 10–18 s Postexperimental Stimulus Onset
Left lateral pad
Right lateral pad
Note. Statistically significant increase in HbO2 or HHb concentration compared to baseline. HbO2 = oxyhemoglobin; HHb = deoxyhemoglobin.
*p < .05, one sample t test.
The grand averaged time courses of the hemodynamic changes in HbO2 and HHb for Channels 14 and 28 are shown in Figure 3. The time courses display the response across infants to the experimental condition followed by the baseline condition. Alongside the significant increase in HbO2 concentration in both channels, the nonsignificant HHb concentration changes vary, with Channel 14 showing a decrease and Channel 28 a slight increase. The latency of the maximum change in HbO2 in both channels occurred at approximately 13–15 s poststimulus onset, toward the end of the trial.
Changes in the concentration of HbO2 showed a significant increase in response to the social dynamic stimuli in two channels located in the posterior area of the pad over each hemisphere, scalp locations T5 and T6 (10–20 system). The location of this activation is consistent with findings of previous fMRI research with adults (for review, see Allison et al., 2000; Pelphrey et al., 2005). The findings support the possibility that 5-month-old infants may already activate a restricted temporal region for the visual processing of social stimuli.
However, an alternative interpretation of the results from Experiment 1 is that the significant response seen in the posterior channels of the lateral pads is attributable to activity in MT/V5 rather than an area of the brain responsible for processing social stimuli such as STS. MT/V5 is a motion-sensitive area of the visual cortex located near to posterior STS, which responds to dynamic stimuli (Henning, Merboldt, & Frahm, 2005; Nogushi, Kaneoke, Kakigi, Tanabe, & Sadato, 2005; Watanabe, Kakigi, Miki, & Puce, 2006). As the experimental stimuli differed from the baseline in both the social nature of the images and in the dynamic motion, we cannot determine which aspect of the dynamic social stimulus elicited the effect observed. Therefore, a second experiment was designed to investigate this question. To explore the importance or otherwise of complex dynamic motion for the activations observed, in Experiment 2 we contrasted two experimental conditions: The first was the same social dynamic video used in the first experiment and the second was a nonsocial dynamic video that involved complex and nested motion patterns. We predict that the posterior temporal effect observed in Experiment 1 would be evident in Experiment 2 for the social dynamic stimuli but not for the nonsocial dynamic stimuli. This result would allow us to support the hypothesis that the observed activation can be attributed to social processing occurring in the STS.
Participants. Twelve 5-month-old infants took part in the second experiment (6 females and 6 males; age = 146 ± 6.4 days old). Another 7 infants participated but were excluded from the study (4 failed to look at the stimuli for the minimum number of trials, 2 became fussy, and in 1, the NIRS headgear was improperly placed). All parents volunteered by responding to advertisements and gave written, informed consent to participate. The study protocol was approved by the appropriate Ethics Committee. The infants were from varied ethnic and socioeconomic backgrounds, predominantly White (British/non-British).
Procedure. The same procedure was used as in Experiment 1 but with two alterations. Experiment 2 compared two experimental conditions: social dynamic and nonsocial dynamic. The social dynamic stimuli were the same as in Experiment 1, while the nonsocial dynamic stimuli were video clips of machine cogs and pistons and moving mechanical toys. These stimuli were selected because they involved complex, interacting curvilinear motion patterns that served as a good control for biological and facial motion. The two experimental conditions were presented sequentially with the baseline trials occurring between each experimental trial (see Figure 1).
In addition, the preferential looking trial was presented at the same two intervals (i.e., one after the sixth trial and one after the 12th trial); however, instead of comparing the social stimulus to baseline, a still image of each of the experimental conditions was displayed, that is, social versus nonsocial.
Data acquisition, array placement, and data rejection. This followed the same methods as Experiment 1 but with one amendment to the data rejection criteria. As there were two types of experimental condition in Experiment 2, infants watched a lower number of trials per condition even though the overall looking time was comparable. Therefore, in Experiment 2, we required a minimum of three valid trials for each experimental condition (social dynamic or nonsocial dynamic).
Data analysis. This followed the same procedure as that used in Experiment 1. In addition to the one-sample t tests performed on each separate experimental condition (social dynamic or nonsocial dynamic vs. baseline), a paired t test was also performed in each channel to compare the activation in response to the two different conditions. As in Experiment 1, these analyses were conducted on the maximum change in HbO2 and HHb compared with baseline, during the specified experimental time window (10–18 s postexperimental stimulus onset).
According to the criteria previously described, we included data from 12 of the 19 infants tested. The mean number of experimental trials recorded in these infants was five social dynamic trials and five nonsocial dynamic trials, and the mean length of the recording session was 6 min, similar to that of Experiment 1. Due to a technical fault in the left lateral pad, Channels 1 and 3 were excluded from seven of the infants’ data, though all of these infants still met the exclusion criteria described earlier. The proportion of valid channels across the group of infants was 67% in the left lateral pad and 77% in the right lateral pad. The preferential looking task indicated that there was a significant preference for the social stimulus over the nonsocial stimulus (dynamic social = 71.6% and nonsocial dynamic = 28.4%, p < .05, t = 4.419, paired t test).
Group effects for the social dynamic experimental condition using a one-sample t test of the maximum increase in HbO2 in response to the stimulus (during the specified time window 10–18 s poststimulus onset) revealed a significant increase from baseline in HbO2 concentration in nine channels (see Table 2). Five of the nine channels were located in the posterior region of the pads in each hemisphere (Channels 11, 14, 25, 27, and 28). Two of these five channels are the same two channels that were found to have a significant increase in HbO2 in response to the social dynamic stimuli in Experiment 1 (Channels 14 and 28). The remaining four channels with significant activation were located in the anterior regions of each lateral pad (Channels 1, 15, 16, and 17). Two of these four channels showed borderline significance in response to the social dynamic stimuli in Experiment 1 (Channels 1 and 16).
Table 2. Group Data of the Statistical Analyses and Mean Maximum Changes in HbO2
Social dynamic experimental condition (one-sample t test)
Left lateral pad
Right lateral pad
Nonsocial dynamic experimental condition (one-sample t test)
Left lateral pad
Right lateral pad
Nonsocial versus social experimental condition (paired-sample t test)
Left lateral pad
Right lateral pad
Note. The results are shown for the channels with a statistically significant increase in HbO2 concentration compared to baseline during the window of activation 10–18 s postexperimental condition onset. Note that for the nonsocial dynamic condition versus baseline, there were no significant results in the left lateral pad; therefore, none are reported. HbO2 = oxyhemoglobin; HHb = deoxyhemoglobin.
*p < .05, one-sample t test.
Group effects for the nonsocial dynamic experimental condition using a one-sample t test of the maximum increase in HbO2 concentration in response to the stimulus (during the specified time window 10–18 s poststimulus onset) revealed a significant increase from the baseline in HbO2 in two channels (Channels 15 and 26; see Table 2). These channels were located in the right lateral pad, one of which located in the anterior region was also found to be active during the social dynamic experimental condition. There were found to be no significant changes in HHb concentration during either experimental condition.
Finally, a paired-sample t test to compare the difference in the maximum increase in HbO2 from baseline during each experimental condition revealed a significant difference between the response for the social dynamic and the nonsocial dynamic stimuli in two channels (see Table 2). These two channels, 14 and 28, displayed a significantly larger increase in HbO2 to the social dynamic stimuli compared with the nonsocial dynamic stimuli. These two channels are the two most posterior channels in each pad and are the same channels that were found to have a significant increase in HbO2 in response to the social dynamic stimuli in Experiment 1.
The grand averaged time courses of the hemodynamic changes in HbO2 and HHb for Channels 14 and 28 (the channels found to show a significant difference in the hemodynamic response across experimental conditions) are shown in Figure 4. The time courses display the response across infants to each experimental condition followed by the baseline condition. During the social dynamic experimental condition, alongside the significant increase in HbO2 concentration in both channels, the nonsignificant HHb concentration changes vary, with Channel 28 (right hemisphere) showing a decrease in HHb and Channel 14 (left hemisphere) first showing a slight increase in HHb followed by a decrease. The latency of the maximum change in HbO2 in both channels occurs at approximately 15–16 s poststimulus onset just as the stimulus is ending, a similar finding to that of Experiment 1. In contrast, during the nonsocial dynamic experimental condition the hemodynamic time courses show two different trends in these two channels. In Channel 14 (left hemisphere), both HbO2 and HHb concentration show a slight increase over the time course, whereas in Channel 28 (right hemisphere), the HbO2 and HHb show the opposite trend to the response observed to the social dynamic stimuli; the HbO2 concentration decreases across the time course and the HHb concentration increases.
Finally, Figure 5 shows a schematic map of the sensor pads for each hemisphere highlighting the significant HbO2 changes seen across infants. A schematic map is shown for each experiment (1 and 2), with separate maps for each statistical comparison (social dynamic vs. baseline [Experiments 1 and 2], nonsocial dynamic vs. baseline [Experiment 2], and social vs. nonsocial [Experiment 2]) illustrating the spatial distribution of the significant responses. This figure highlights the clustered effect in the posterior and anterior regions seen in response to the social dynamic stimuli (Figure 5b).
The findings of Experiment 2 reveal several significant effects in support of our hypotheses. First, there were significant hemodynamic increases in HbO2 in response to the social dynamic stimuli in the posterior region of each sensor pad, localized to scalp locations T5 and T6 (10/20 system). These effects replicate those obtained in Experiment 1. Additionally, a consideration to be addressed by Experiment 2 was whether the posterior hemodynamic change observed in response to the social dynamic stimuli could be attributed to MT/V5, a general motion sensitive area, rather than attributed to a social stimulus specific response in the STS.
In Experiment 2, the nonsocial dynamic stimuli did not reveal any significant HbO2 concentration changes in the posterior channels that showed significant effects in response to the dynamic social stimuli in Experiments 1 and 2. Further, when the response to the dynamic social stimuli was directly compared with that to the nonsocial dynamic stimuli, a significant difference was found between the two conditions in bilateral posterior channels. The social dynamic stimuli had a significantly greater increase in HbO2 compared to the nonsocial dynamic stimuli in the most posterior channels of the pads near to scalp locations T5 and T6 (10/20 system). These findings provide further support for the hypothesis that infants as young as 5 months of age have a specialized area of the temporal cortex activated by dynamic social stimuli. The location of this neural activity identified by the significant hemodynamic changes is consistent with the findings of fMRI studies on STS activity in adults in response to social stimuli (for review, see Allison et al., 2000; Pelphrey et al., 2005).
An unpredicted region of activation in response to social dynamic stimuli was also found in the most anterior area of the pad over each hemisphere, with grand averaged HbO2 hemodynamic changes approaching significance in Experiment 1 and exhibiting significant increases in response to the social dynamic stimuli in Experiment 2. Moreover, a significant increase in HbO2 was also found in one of these significant right anterior channels in response to the nonsocial dynamic stimuli, indicating a more general response property.
We have used optical topography to detect differences in stimulus processing in 5-month-old infants. Our results demonstrate greater neural activation, as measured by relative changes in the hemodynamic response, in regions of the posterior lateral pads in response to dynamic social stimuli. This area corresponds approximately to scalp locations T5 and T6, which are located over the region of the STS where neural activation is found in adults in response to social stimuli. These findings provide further support for the hypothesis that infants as young as 5 months of age have a specialized area of the temporal cortex for processing social dynamic stimuli. In further support of this hypothesis, neural activation as measured by hemodynamic change in HbO2 to the nonsocial dynamic stimuli, was not evident in the posterior region of the pads. These findings are in accord with an adult fMRI study by Beauchamp, Lee, Haxby, and Martin (2002) who observed a greater response in the superior temporal cortex to moving human stimuli compared with moving tools.
Interestingly, a more widespread response to the social dynamic stimuli was observed in Experiment 2 compared to Experiment 1 (i.e., Experiment 2 had a greater number of channels with significant hemodynamic changes). One difference between the first and second experiments was the number of successful trials that the infants attended to, reducing in number from an average of 10 trials in Experiment 1 to 5 trials in Experiment 2. This was an unavoidable consequence of adding a second experimental condition. Is it possible that as the stimulus presentation continues, the infant’s neural responses adapt to the stimuli becoming reduced in amplitude over time? The adaptation of neural responses to repeated stimulus events has been identified using a number of measures from single unit recording to fMRI (Krekelberg, Boynton, & van Wezel, 2006). Taking this into consideration, we reran the group analyses for Experiment 1, including only the first 5 trials from each infant. We found that the number of significant channels increased from Channels 14 and 28, to Channels 14, 25, and 28 in the posterior region and Channels 2, 15, and 17 in the anterior region of the lateral pads. This finding suggests that infants’ neural responses do indeed adapt to the stimuli, as has been found in fMRI studies with adults (several mechanisms have been proposed for this effect—habituation leading to suppression of temporally repetitive stimuli, Grill-Spector & Malach, 2001; priming of selective neurons, Schacter & Buckner, 1998; and inhibition of nonselective neurons, Li, Miller & Desimone, 1993), and/or that the response lowers as infant’s become less attentive to the stimuli. Interestingly, this effect was found to not occur as strongly in the posterior areas of the lateral pads, as significant channels were found in Experiment 1 over the full set of trials.
The anterior activation as discussed earlier occurred extensively in response to the social dynamic stimuli (i.e., four channels, left and right hemisphere) and to a lesser extent in response to the nonsocial dynamic stimuli (one channel in right hemisphere) in a region approximately between scalp locations T4, F4, and F8. There are several cognitive processes and associated neural activity evident in adults that could provide tentative interpretations for this anterior cortical activation. Although research into neural activation in response to biological motion and social dynamic stimuli has primarily focused on posterior temporal areas such as STS, often additional activation is found in an anterior temporal/inferior frontal region of the cortex. For example, a study by Saygin, Wilson, Hagler, Bates, and Sereno (2004) of biological motion perceived in point light displays found bilateral activation in the inferior frontal gyrus. This activation is in a similar area to the region of anterior activation found in response to the social dynamic stimuli in the current study. The authors conclude that this frontal activity reflects the recruitment of action observation networks and provides a possible interpretation for the anterior location of activity in this study. Further, a meta-analysis of functional neuroimaging studies (Pelphrey et al., 2005) of biological motion suggests that mouth movements are often associated with the anterior temporal and/or left area of the STS, while eye gaze is often associated with the posterior temporal and/or right area of the STS. Moreover in an fMRI study conducted to support this meta-analysis, Pelphrey et al. (2005) also found that mouth and eye movements (but not hand movements) caused bilateral activation in the middle and inferior frontal gyrus. It is evident that perception of differing social movements can cause neural activation in different locations of the cortex in adults. Although the current study used a variety of social movements in the video clips (hand, eye, and mouth) potentially allowing us to investigate such effects, the clips were mixed together during each 16-s trial and so separate hemodynamic responses for each type of social movement could not be assessed. It is clear that further developmental research is needed to investigate whether the spatial localization of the hemodynamic responses observed in this study could potentially differ across varying social actions, or whether the localized areas of neural activation reported in adult research will not yet be evident in infancy.
Although far greater anterior neural activation was observed in response to the social stimuli, some effect was also seen in response to the nonsocial dynamic stimuli, which was restricted to the right hemisphere. Research on spatial working memory as evidenced in a recent study by Imaruoka, Saiki, and Miyauchi (2005) may offer a further explanation for this neural activity. These authors investigated neural responses to a multiple component object permanence-tracking task in which participants were required to maintain a coherent percept of a multiple component object either while it was stationary or dynamic. During the dynamic condition there was significant activation in the ventral-frontal cortex strongly lateralized to the right hemisphere, a region of the cortex similar to the activation found in the current study to the nonsocial and social dynamic stimuli. In support of this finding, a meta-analysis by D’Esposito et al. (1998) concluded that activity in the ventral-frontal areas in response to spatial working memory tasks is often strongly lateralized to the right hemisphere. Further to this, the ventral-frontal cortex including the inferior frontal gyrus is within the frontoparietal network, which is often associated with a range of diverse cognitive demands including goal-directed and stimulus-driven attention (Corbetta & Shulman, 2002; Culham et al., 1998;Duncan & Owen, 2000; Jovicich et al., 2001).
As to our knowledge there are currently no other infant neuroimaging studies of this nature, we have to rely on research with adults to provide support or an explanation for our findings. Potentially, infants could be responding to the multicomponent dynamic objects used in the current study in the same way that adults do during such processes as spatial working memory, goal-directed and stimulus-driven attention, or action observation networks (for the social dynamic stimuli). Further developmental research is needed to investigate whether the ventral-frontal spatial localization of the hemodynamic responses observed in this study could be attributed to any of the processes outlined above.
We have demonstrated that young infants may already show activation of a specific area of the temporal cortex in response to social stimuli, evidenced by localized HbO2 responses in the posterior channels of the temporal pads bilaterally. We were, however, unable to detect any significant decreases in HHb in response to the social stimuli. While adult NIRS studies show HbO2 and HHb responses consistent with BOLD fMRI signal changes (i.e., a decrease in HHb and parallel increase in HbO2), infant data reveal a less consistent pattern of HHb changes in response to neural activation (Baird et al., 2002; Meek et al., 1998; Sakatani et al., 1999; Wilcox et al., 2005). It is uncertain whether in infants the standard decrease in HHb should be expected, and is a matter of much controversy. Related to this, there is also ongoing debate concerning the ideal pair of wavelengths to use for NIRS measurements (Boas, Dale, & Franceschini, 2004; Sato, Kiguchi, Kawaguchi, & Maki, 2004; Yamashita, Maki, & Koizumi, 2001) and, in particular, how the wavelengths of the source lasers may effect the accurate measurement of HHb changes. Indeed, it is possible that the wavelengths of the source lasers in the multichannel system used in this study may not be as sensitive to HHb changes as to the estimation of HbO2. We are currently investigating this wavelength-dependent response and this debate is being addressed in future generations of NIRS systems.
Despite these technical issues, we can clearly measure HbO2 hemodynamic responses to social and nonsocial dynamic stimuli. Further work with this technique may help us understand whether infants display differential cortical responses to differing types of social stimuli and whether we can associate this activity with the STS, as has been shown in fMRI studies with adults.
Does the activation of posterior temporal regions to social dynamic stimuli in 5-month-old infants imply the early specialization of the cortical social brain network? While other studies are consistent with the rapid specialization of regions of the cortical social brain network (Grossmann et al., 2008), a large body of electrophysiological and fMRI data indicate relatively gradual specialization of key components of the adult social brain network, such as the fusiform face area (e.g., Cohen Kadosh & Johnson, 2007; de Haan, Pascalis, & Johnson, 2002). Although the degree of specialization of the posterior temporal areas remains to be investigated in more detail, one possibility is that regions of the social brain critical for detecting communicative situations specialize before other aspects of social brain function (Grossmann & Johnson, 2007). The findings of this study highlight the exciting potential of optical topography as a tool for studying these critical issues in developmental cognitive neuroscience.