Looking Time Data
The infants' looking times during the test trials were averaged across trial and infant for each condition (multi-featural change, M = 28.04, SD = 1.86; shape change, M = 27.81, SD = 1.47, color change, M = 27.64, SD = 2.37; control, M = 26.84, SD = 1.62). The infants in all four conditions looked almost continuously throughout the test trials, suggesting that they found all four events engaging.
The hemoglobin concentration response curves are shown in . Given evidence that HbO2
is a more robust and reliable measure of neural activation than HbR or HbT (Bartocci et al., 2001; Chen et al., 2002
; Hoshi &Tamura, 1993
; Jasdzewski et al., 2003
; Kato et al., 1993
; Sakatani et al., 1999
; Strangman et al., 2002
), we focused on HbO2
in our statistical analyses. For comparison purposes, however, the mean HbO2
, HbR and HbT responses are reported in . Relative changes in HbO2
concentration from 5 to 30 s following initiation of the event (prior to 5 s the hemodynamic response is still being initiated) were compared to the baseline from -1 to 0 s (). As predicted, the infants in all four conditions demonstrated a significant increase in HbO2
in occipital cortex in response to the occlusion event (see figure caption). A one-way analysis of variance (ANOVA) was conducted with condition as the between-subjects factor. The effect of condition was not significant, F
(1, 31) < 1. Planned comparisons revealed that the mean HbO2
response observed in the control condition did not differ reliably from that observed in the multi-featural, shape, and color change conditions, all F
s(1,31) < 1. Although qualitative inspection of the data suggests that the response magnitude observed in the color change condition was lower than that observed in the other three conditions, this outcome was not predicted nor did it reach significance. Follow-up research will be needed to establish the extent to which this outcome is reliably observed in object processing tasks.
Figure 3 Hemoglobin response curves averaged across subjects and trials (with SE bars) for the occipital and inferior temporal cortices. Each row shows the response curves, in optical density units, for the indicated condition. The event began at time 0 and continued (more ...)
Relative changes in HbO2, HbR and HbT concentration following initiation of the test event, by neural region and condition. Cells contain M (SE) optical density units averaged from 5 to 30 s.
Figure 4 The mean (and SE) oxyhemoglobin (HbO2) response for each condition and brain region, reported in optical density units. Responses were compared to a mean value of 0. The ‘*” indicates those responses that are significantly different from (more ...)
A different pattern of responses were observed in IT (). The infants in the multi-featural and the shape change condition evidenced a significant increase in HbO2 (). The infants in the color change condition also evidenced an increase in HbO2 but this increase did not differ significantly from 0. In contrast, the infants in the control condition evidenced a small, non-significant decrease in HbO2. A one-way ANOVA did not yield a significant effect of condition, F(1, 31) = 2.06, p = .127, ŋp2 = .17. However, planned comparisons indicated that the hemodynamic response observed in the control condition differed from that observed in the shape condition, F(1,31) = 4.54, p = .041, and the multi-featural condition, F(1,31) = 4.10, p = .052, although the latter was only marginally significant. In contrast, the hemodynamic response observed in the control condition did not differ reliably from that observed in the color condition, F(1,31) = 3.40, p = .075. A final comparison revealed that that the hemodynamic response observed in the shape condition did not differ significantly from that observed in the color condition, F (1,31) < 1.
Together, these results suggest several conclusions. First, neural activation, as measured by the hemodynamic response, is observed in O in response to occlusion events involving objects, regardless of whether the objects seen to each side of the screen are identical or vary on one or more feature dimensions. Second, the degree of neural activation observed in IT depends on the feature dimension manipulated. When the objects seen to each side of the screen differ on many dimensions or on shape only, a significant increase in neural activation is observed and these responses differ from those observed when the same object is seen to both sides of the screen. When the objects differ in color only, an increase in neural activation is observed but this increase does not differ significantly from baseline. Qualitative inspection of the data suggests that a color change does lead to some degree of activation in IT as compared to no change, but that the magnitude of the response is not as great as that observed in response to a shape change. However, statistical tests comparing the responses of the infants in the color change condition to those of the infants in the shape change and the control condition were not significant.
This outcome leads to the third conclusion, which is that between-subjects' designs do not offer sufficient power to test the type of hypotheses we proposed. We employed a between-subjects design because this is the design we typically use in our behavioral studies, with sample sizes similar to those reported here, and we typically have sufficient power to test for differences between conditions. However, the NIRS data contain greater between-subject variability than that typically observed in our behavioral work. We anticipate that as we improve on probe and headband design the quality of the data will improve, resulting in greater statistical power. At the same time, there are some factors over which we have little control that also contribute to between-subject variability. For example, given what we know about the structural organization of the adult brain, we expect that there are individual differences in the size and location of the neural regions under study. Although we use the most reliable system currently available for probe positioning, there is no guarantee that the location of the underlying cortical structures, in relation to external (e.g., skull) markers is the same for each infant. Until more accurate methods become available, we plan to continue to use the 10-system and employ within-subjects designs to decrease one source of variability.