Two Brains, One Child: Interhemispheric Information Transfer Deficits and Confabulation in Children Aged 4, 7, & 10 (Reprinted from Cortex, 20, 317-331, 1984).
by Rhawn Gabriel Joseph, et al.
Rhawn Gabriel Joseph, Ph.D.
R.E. Gallagher, Wendy Holloway and Judy Kahn
UHS/The Chicago Medical School
In an investigation of developmental trends in the ability to transfer information between the cerebral hemispheres, 138 children, ages 4, 7, 10, viewed pictures presented to either the right or left hemisphere and were asked to describe what was viewed. Responses were scored for the number of accurate pictorial features reported (inclusion score), confabulation (erroneous embellishment), perceptual, semantic, and syntactic errors. Misses (failure to provide a response) were also noted. A preponderance of errors were found, such that 4-year olds engaged in a significant degree of confabulatory responding following right vs left presentation. A significant inverse relationship was found between inclusion and confabulation such that the larger the gap in information reported the greater was the tendency to insert (confabulate) erroneous material. In addition, all children were given a tactile-shaped recognition task employing both transfer and non-transfer conditions and requiring tactile exploration and visual recognition. Significantly more errors were found on the transfer vs non transfer task among the 4-year olds only. In that deficits in information transfer were not complete but only partial, and given that significant intra-hemispheric (non-transfer) errors were also discovered, it was concluded that the hemispheres of children age 4 are not completely but only partially disconnected, and that intra-hemispheric (cortical) immaturity plays a significant role in the production of processing deficits.
A considerable body of evidence has been presented showing that traumatic or surgical section of the corpus callosum results in interhemispheric deficits in information transfer (Bogen, 1979; Gazzaniga and LeDoux, 1978; Geschwind, 1965; Sperry, 1974). A number of investigators have provided evidence which purports to show that young children have similar difficulties in interhemispheric information transfer (Finlayson, 1976; Galin, Diamond and Herron, 1977; Galin, Johnstone, Nakell and Herron, 1979; Gallagher and Joseph, 1982; Joseph, Lesevre and Dreyfus-Brisac, 1976; Kraft, Mitchell, Languis and Wheatley, 1980; O'Leary, 1980; Salamy, 1978). These deficits, presumably, are secondary to the prolonged development (e.g. myelination) of the callosum, the maturation of which is not complete until after the first decade (Yakovlev and Lecours, 1967). Hence, the capacity to transfer information between the cerebral hemispheres appears to progressively increase with age (Finlayson, 1976; Galin et al., 1977; Kraft et al., 1980; O'Leary, 1980).
Consonant with the behavioral evidence cited above are some electrophysiological studies. Briefly, Salamy (1978) failed to find ipsilateral evoked potentials (which are perhaps dependent on transfer via the corpus callosum) prior to age 3.5, while Joseph et al. (1976) found deficiencies in the spatio-temporal organization of the EEG and thus a general lack of interhemispheric linkage in newborn and premature infants.
It is important to note, however, that in the behavioral studies, some transfer of information occurred in all cases, and some children performed without errors. Moreover, Finlayson (1976) required transfer only from the right to left (untrained) hand; Galin et al. (1979) employed only female subjects, and O'Leary (1980) failed to find transfer deficits on two of the four tasks utilized. In addition, the results of Salamy (1978), and in particular Gallagher and Joseph (1982) and Kraft et al. (1980) may in part be a function of parietal lobe immaturity.
The evidence, thought far from conclusive, is thus indicative of limited rather than absent interhemispheric information exchange in young children. Indeed, given the differing local rates of cortical and callosal fiber development there is likely to be considerable transfer depending on age and modalities employed. Furthermore, since the anterior commissure (a tract of fibers which interconnects the inferior temporal lobes and limbic areas) appears developmentally complete by the end of the first year of life, even the hemispheres of the very young child may be only partially disconnected.
Confabulation and Limited Interhemispheric Information Exchange
As noted by Bogen (1979), Sperry (1974) and colleagues, in cases of complete section of the commissures each hemisphere appears ignorant of the cognitive activities occurring in the other. In cases of partial sectioning (e.g. sparing of the anterior commissure) some transfer occurs (e.g. Gazzaniga and colleagues, 1978, 1979), albeit limited and incomplete. For example, in one patient (PC) with an intact anterior commissure, following right hemisphere tachistoscopic pictorial stimulation all verbal reports, while containing some accurate descriptive elements, were largely contaminated with errors, gross embellishments, and other falsehoods. Information transfer (from the right hemisphere) was thus characterized by confabulatory "gap-filling" (cf. Geschwind, 1965; Joseph, 1982; Talland; 1961), perhaps because the left hemisphere apparently attempted to make sense of the limited information received.
Thus, in the case presented by Gazzaniga and colleagues (1978, 1979) when a complex picture, e.g. a man with a gun, was presented tachistoscopically to the patient's left hemisphere, he would typically respond with concrete and accurate two or three word replies, such as "guy with a gun". When the same stimulus was presented to the right hemisphere (thus requiring transfer to the language centers of the left) he would respond with elaborate, detailed, and vivid descriptions which bore little relation to the actual elements of the original stimulus around which other associations were erroneously anchored: "Gun, hold up╔he has a gun and is holding up a bank teller, a counter separates them." Furthermore, this patient's left hemisphere apparently believed the largely erroneous descriptions offered in these instances, for, as noted by Gazzaniga and LeDoux (1978), it did not "offer its suggestion in a guessing vein but rather as a statement of fact" (pp. 148-149).
The results of Gazzaniga and colleagues (1978, 1979), when considered in light of the possibly limited information exchange between the hemispheres of children, suggest that young children might also confabulate if questioned concerning pictorial stimuli presented to the right hemisphere.
Confabulation. Anchors and Gaps.
It has been argued (Joseph, 1982) that confabulation may occur when the language centers of the left hemisphere are isolated from sources of information about which the patient is questioned, such that complete and accurate information is not fully available. In these instances confabulation may occur as a result of attempts to fill the gaps in the information received with associations which are in some manner related to the fragments available (Joseph, 1982; Talland, 1961). Moreover, disconnection also results in failure to correct erroneous statements, as contradictory evidence is also not always accessible. Hence, confabulatory statements, although erroneous, often contain accurate elements around which erroneous (albeit related) ideations are anchored. However, confabulation may also result due to other factors such as frontal lobe dysfunction such that (in these instances) an uninhibited flow of irrelevant information is transmitted to and organized by the language regions.
MATERIAL AND METHOD
Three age groups of children (4, 7, 10) were tested and asked to describe pictorial stimuli presented to the left or right cerebral hemisphere. It was predicted that due to the immature state of the corpus callosum and the cerebral cortex, children would incompletely transfer, inaccurately report and embellish information presented to the right as compared to the left hemisphere. These subjects were also tested on the Tactile Form Recognition task of the Halstead-Reitan Battery, and their tendency to make transfer vs. non-transfer error ascertained.
For the purposes of this study confabulation was defined as an erroneous embellished description, such that perceptual-pictorial information not originally presented was inaccurately inserted and included in the subject's report (cf. Exner, 1974; Klopfer, Ainsworth, Klopfer and Holt, 1954; Talland, 1961). Hence, a detail may be reported correctly but the meaning of the detail may become overly significant and erroneously generalized such that a concept results which has little or, in fact, no bearing on the original stimulus.
Information Gaps and Inclusion Score
An important feature of the gap filling hypothesis is the notion that information degradation and confabulation should be directly related, such that, the larger the gap, the more likely is the individual to confabulate. Hence, for purposes of this study the major pictorial-perceptual elements (features) of each tachistoscopic stimulus were determined. For example, if in response to Stimulus J (see Figure 1): a girl blowing out the candles on a birthday cake, a subject responded, "girl" or "cake", the inclusion score would be "I" (as opposed to "2" for stating "girl" and "cake"). For each presentation the inclusion of these items in the child's verbal report was therefore tabulated.
Misses, Semantic, Syntactic, and Perceptual Errors
Tabulation of the frequency of misses, semantic, syntactic and perceptual errors was largely for exploratory purposes. Briefly, misses were scored when no response was offered or the subjects claimed not to have seen the stimulus. Syntactic errors were noted when the subject's speech was aggrammatical. A response was scored as a semantic error when the words used in the description, although incorrect, were from the correctly aroused semantic group pertaining to the stimulus (cf. Lhermitte and Beauvois, 1966). The erroneous response therefore bore a meaningful resemblance to the original stimulus. In contrast, a perceptual error was tabulated when an erroneous response/description bore an obvious similarity to some aspect of the stimulus whole (see Figure 1); for example, calling, calling a "cake" a "drum". However, unlike a confabulation, no erroneous information could be added (for either a semantic or perceptual error). In this respect it was hypothesized that a perceptual error probably would result from a breakdown of serial feature-by-feature visual scene analysis as well as a failure to adequately engage in visual exploration (see Rubens, 1979, regarding agnosia). Thus, only certain details are recognized in isolation and stripped of related features. Hence, the left hemisphere should perform more poorly.
All subjects were right handed, aged 4, 7, 10, predominantly white, with approximately 12% of each group consisting of minority children (e.g. Black, Asian). A total of 132 subjects were tested, with a minimum of 20 girls and 20 boys in each age group (4, N=48; 7, N=44; 10, N=40). Subjects were volunteers recruited from private schools and were obtained via parents' consent.
All children were tested in an isolated and quiet room provided by the school. For assessing tendencies to confabulate or make other errors in description, a single-field tachistoscope was employed and all responses were tape-recorded. Visual stimuli consisted of pictures adopted from the Peabody Picture Vocabulary Test (PPVT; Dunn, 1959). Eleven stimuli were employed for presentation to either the right, left, or visual-field midline, for a total of 33 presentations. In addition, two control stimuli and a pre-test "warm-up" stimulus from the PPVT were used. All stimuli were projected on a small white screen with a single fixation point (1.54 cm hole) at its center.
For the tactile-form recognition test, the standard Halstead-Reitan Form Board was employed. Briefly, this is a wooden rectangular box which contains a hole for hand insertion at is bottom front center. Across the face of the apparatus are embedded four small plastic chips, each of which is identical to one of the four stimulus objects (i.e. cross, triangle, square, circle) used for non-visual tactile exploration.
Procedure for Tactile-Form Recognition Testing
Prior to testing each child was questioned about and asked to demonstrate (on 3 tasks) hand usage. Inclusion criterion was right handedness on 2 of the 3 tasks. After being familiarized with the stimuli and procedures the subject was instructed to place his right hand through the hole of the apparatus. A single chip was placed in the fingers and the subject encouraged to feel it until he knew what it was. The subjects were then instructed to release the chip, remove his hand, and thus use the same hand (tactile-visual non-transfer, TVNT), to point to the object on the face of the form board. After 4 trials, once with each chip, the procedure for non-transfer recognition was repeated for the left hand.
Instructions for the transfer condition were similar except that the child was asked to point with the opposite hand (tactile-visual transfer, TVT) to the correct object. After this procedure was repeated once for each object, the child was instructed to insert the left hand and the transfer trials were repeated.
Testing always began with the right hand. However, for approximately half of the subjects at each age level, testing began with the non-transfer trials, and with the transfer trials for the remaining subjects.
Procedure for Tachistoscopic Testing
Following the Tactile Form Recognition tests, each subject was presented with a picture from the PPVT and encouraged to give a detailed description of it. The child was then seated facing the viewing screen, asked to maintain visual fixation on the center point even when the picture fell to either side, and asked to describe each picture presented. Two practice trials in which images were presented at the screen's center were run and the child encouraged to describe what was seen. Testing was then initiated and 33 trials were run.
All stimuli were presented once to each visual field, the right or left hand edge at least 3 degrees from the central fixation point, for exactly 150 msec., and last to the screen's midline. Left and right presentations were counter-balanced (i.e. stimuli presented initially to the right then left and midline, 5 different stimuli initially to the left then right and midline, and one stimulus presented initially the midline followed by left then right sided presentations). All slides were presented in a pre-established random order which remained constant for all subjects. Eye movements were carefully monitored by an observer hidden behind the viewing screen, via the center-fixation hole.
After each presentation the subject was asked, "What did you see?" However, if the child's response was a one word reply (e.g. "girl"), the experimenter asked "what else did you see?" If the child again responded with a one word description the next stimulus was presented and the procedure repeated.
Scoring and Rater Reliability
Three raters, blind to mode of presentation (i.e. right, left, midline) were given intensive instruction and training in scoring procedures and criteria for the tachistoscopic responses, and extensive practice in the scoring of protocols. Inter-rater reliability, based on the scoring of 8 research protocols, yielded reliability coefficients in excess of .90.
Significant main effects were found for age (Fw96.20, p< .00001) and mode of visual-field presentation (Fw108.94, p<.00001). A Newman-Keuls (NK) analysis of means indicated that 4-year olds had significantly lower inclusion scores (summed across presentation modes) than 7 (Qw16.31, p<.01) or 10-years old children (Qw22.68, p<.01), and that subjects age 7 responded with significantly fewer inclusion items than the 10-year old group (Qw6.37, p<.01). In addition, across all age groups inclusion scores were found to be significantly larger for left hemispheres presentation than for right (Qw1.202, p<.01). Midline exposures also resulted in a significantly grater number of inclusion elements than left (Qw2.47, p<.01) or right presentation (Qw3.672, p<.01). However, this latter finding is probably due to having viewed all stimuli two times prior to midline presentations (once right and once left). The NK analysis of experimenter influences did not reveal any significant main effects.
The ANOVA demonstrated a significant effect for age (Fw26.25, p<.00001) and visual field presentation (Fw64.37, p<.00001). The NK analysis on ordered pairs of group means indicated that 4-year olds made significantly more confabulatory responses that 7 (Qw4.24, p<.01) or 10-year olds (Qw5.383, p<.01). The age 7 group also confabulated significantly more than the older children (Qw1.04, p<.05). In addition, more confabulatory responses were found to occur following right vs. left hemisphere presentation (Qw.803, p<.01), or midline exposures (Qw.834, p<.01).
A significant interaction between presentation site and age was also indicated (Fw4.26, p<.002), such that regardless of right hemisphere (RH), left hemisphere (LH) or midline (MH) presentation, 4-year olds confabulated significantly more than 7 (RH: Qw2.133, p<.01; LH: Q<.932, p<.01; MH: Qw1.178; p<.01) or 10-year old children (RH: Qw2.492, p<.01; HM: Qw1.375, p<.01; MH, Qw1.517, p<.01). In addition, among the children age 4, there were significantly more confabulatory responses to the right vs. left (Qw1.542, p<.01) or midline presentation (Qw2.25, p<.01). Among 7 and 10-year olds no significant age or presentation mode differences were discovered. No significant main effects for experimenter influences were uncovered.
Semantic and Syntactic Errors and Misses
No significant main effects for syntactic errors were found. A main effect for semantic errors was discovered (Fw9.10, p<.001), such that 10-year old subjects made fewer errors than 4 (Qw2.66, pw.01) or 7-year old children (Qw2.386, p<.01). Children age 4 and 7 did not significantly differ.
Significant differences in the tendency to omit responses (Misses) were found to be age dependent (Fw8.86, p<.0003). It was determined that 4-year olds had significantly more misses than 7 (Qw.602, p<.01) or 10-year olds (Qw.775, p<.01). The 7 and 10-year groups were not found to significantly differ.
Significant age-related differences in the tendency to commit perceptual errors were noted by the ANOVA (Fw5.67, p<.005). The NK analysis indicated that 4-year old children made significantly more perceptual errors than 7 (Qw.442, p<.01) or 10-year olds (Qw.994, p<.01). In addition, children age 7 committed more perceptual errors than older group (Qw.554, p<.01).
In regard to overall scores summed across age and mode of presentation, it was also noted that the experimenters were associated with significantly different amounts of perceptual errors (Fw7.33, p<.0002), such that Experimenters (E) 1 and 3 significantly differed from each other and from E2 and E4 (p<.01), who in turn did not differ from each other (Qw.12, pwNS). A check of the raw data for this and all other effects associated with the different subjects tested by each experimenter indicated, however, that these differences are largely a function random subject variations (probably associated with demographic characteristics- e.g. socio-environmental influences) and the tendency of a few subjects to respond with an unusually large number of errors.
When considering total error scores for both the transfer (TVT) and non-transfer conditions (TVNT), significant differences were found across age groups (Fw8.86, p<.0003), such that the 4-year group made more errors overall than did 7 (Qw.6818, p<.01) or 10-year olds (Qw.7375, p<.01). Also, regardless of age, significantly more transfer vs non-transfer errors were discovered (Fw13.11, p<.0004). In addition, the ANOVA showed a significant interaction effect for task and age (Fw7.94, p<.0006). In this regard, 4's made significantly more transfer errors than either 7 (Qw.489; p<.05) or 10's (Qw.575, p<.01), and 4-year olds, unlike older subjects, were found to perform more poorly on the transfer vs non-transfer tactile conditions (Qw.4375, p<.05).
Correlations Among Tachiscopic and Tactile Conditions
A multiple correlational analysis (SCSS) demonstrated, as predicted, a highly significant inverse relationship between inclusion scores (i.e. pictorial elements accurately reported) and confabulation (rwy.484, p<.001) across modes of hemispheric presentation. Hence, the fewer pictorial elements reported, the more likely was there a tendency to confabulate. In addition, age was highly correlated with inclusion scores (rw.764, p<.001) and confabulatory responding (rwy.530, p<.005), confirming that with increasing age the number of pictorial features reported (inclusion score) increases, whereas the tendency to confabulate significantly diminishes.
Interestingly, both TVT (transfer) and TVNT (non-transfer) tactile-recognition errors were found to be positively associated with confabulatory responding (rw.215, p<.01; rw.192, p<.05, respectively), and negatively linked with inclusion scores (TVT: rwy.336, p<.01; TVNT: rwy.158, p<.05), with the overall magnitude of correlation for both confabulation and inclusion score being grater for transfer errors.
For example, in response to the tachistoscopic presentation of a picture depicting a girl blowing out the candles on a birthday cake (i.e. stimulus J, Figure 1), many young children fail to report a number of major pictorial features (such as the girl or cake), and thereby earn a low inclusion score. Although after right hemisphere stimulation some elements are accurately reported (as determined by inclusion score), there is a significant failure to account for a number of major perceptual-pictorial elements as compared to the left hemisphere. It thus appears that some information is deleted or degraded during transfer.
In addition, a large number of children were found to erroneously extrapolate and to embellish their descriptions (i.e. confabulate), particularly when pictorial stimuli were tachistoscopically presented to the right hemisphere. This was especially evident among children age 4, as right hemisphere confabulatory errors greatly exceeded those of the left. In contrast, the transfer (i.e. from the right to left hemisphere) and non-transfer (left hemisphere presentation) confabulation scores of the older children were significantly fewer in number and did not significantly differ.
Interestingly, the lower the inclusion score (the fewer items reported), the more likely was a subject to confabulate. This latter significant inverse correlation, maintained regardless of age or presentation field, was particularly strong (see Table I) following right hemisphere stimulation and thus information transfer to the left cerebral cortex. Thus, in conjunction with information deletion (i.e. low inclusion score), erroneous information was somehow added to the subject's report, such that the fragments received became subsequently embellished. For example, in response to stimulus J (girl blowing out the candles on a cake), the child might respond: "a girl opening a package", or "a girl eating something", and thus be scored as confabulatory. In both instances, features such as cake, candles, girl blowing, are deleted and information such as opening a package, or eating something, are inserted.
Why didn't the child simply state, "I don't know" rather than confabulate when confronted by these gaps? We presume this response was rare because that information was not available to the language axis of the left hemisphere. According to the model discussed at length elsewhere (Joseph, 1982) confabulation in part results in response to gaps in information (see also Geschwind, 1965; Talland, 1961), gaps which in turn appear to be due to the transmission of information which is degraded or incomplete when received by Wernicke's region. This type of confabulation is thus the result of the language center's attempts to organize and make sense of the limited information received, by insertion of additional information. The process of insertion is not random, but is partly determined by contextual variables (e.g. emotional status, past history; see Sperry, Zaidel and Zaidel, 1979), as well as the availability of associations/ideations which are linked to the perceived fragments of information. Essentially, the language axis of the left hemisphere does not know that it does not know as it attempts to make sense and report only what is transmitted to it.
It is noteworthy that unilateral left cerebral presentation also resulted in confabulation. This suggests that intra-hemispheric substrate immaturity as well as that of the corpus callosum may also be responsible for part of the observed findings. One candidate for the site of intra-hemispheric deficiencies is the inferior parietal lobule. This area is one of the last to myelinate (Flechsig, 1901), the last in which dendrites appear, and at age 4 is only 64% developmentally complete (Blinkov and Glezer, 1968). Fiber interconnections linking this with other neocortical areas are also not completely matured until approximately 7 years of age (Lecours 1975). In that the inferior parietal lobule presumably has a major role in information assimilation and transfer to Wernicke's region, the lack of maturation in this area may contribute significantly to the observed deficiencies.
Significant differences in transfer vs. non-transfer were not noted for misses (i.e. trails in which no response was elicited), perceptual, semantic, or syntactic errors. In part, the failure to discover transfer effects on these processes may be due to greater efficiency and degree of maturation in the anatomical systems involved, or inadequacies within the methodology. However, the former explanation does not seem completely valid as significant age-related differences were apparent. It is noteworthy that misses, confabulation, and perceptual and semantic errors were significantly large in number regardless of hemisphere of presentation. That is, significant developmental increases in the ability to process and express select forms of information was apparent on almost every measure employed including TVNT. In part, these findings bolster the suggestion advanced above, that possible intra-hemispheric age-related deficits/immaturities may be responsible for these errors in addition to (and perhaps independent of) corpus callosum immaturity.
Indeed, intra-hemispheric deficits in processing may be a significant factor which differentially influenced the overall results. That is, the tendency to make errors in either the right or left hemisphere independent of transfer, may, when transfer is required, give rise to errors in the right and left hemisphere and thus a larger error score independent of and/or addition to the corpus callosum.
On the other hand, particularly as related to semantic and perceptual errors, correct and efficient visual, tactile, and linguistic identification may require the dual and shared (cooperative) activity of both cerebral hemispheres. Hence, processing deficits within each hemisphere may due to an inability to gain complete access to perceptual activity occurring in the other. Assuming the mature callosum does not merely relay information but acts to mediate interhemispheric information integration by allowing for the participation and unique contribution of each hemisphere to the analysis of information, the immature callosum may in effect prevent or decrease each hemispheres, ability to contribute to and integrate the analysis being performed by the other. As such, unilateral processing may suffer as much as operations requiring transfer from one hemisphere to the other.