Implicit memory is independent from IQ and age but not from etiology: evidence from Down and Williams syndromes

Introduction

Implicit memory is manifested as a facilitation (i.e. performance improvement) in perceptual, cognitive and motor tasks, without any conscious reference to previous experiences (Squire 1987); for a review see Schacter et al. (2004). In patients with organic amnesia, the neuropsychological dissociation between poor declarative memory but normal performance on a variety of implicit memory tasks documents the independence of the neurobiological substrates underlying the two forms of memory (Squire et al. 2004). Squire et al. (Squire 1987; Squire et al. 2004) also suggested that in the context of implicit memory, distinct neural circuits are implicated in repetition priming, in learning visual-motor procedures, and in operant conditioning. This additional fragmentation is also supported by neuropsychological evidence that some brain disorders (i.e. Alzheimer's disease) impair repetition priming without affecting procedural learning, while other pathological conditions (i.e. Huntington's disease) compromise procedural learning but leave repetition priming intact (Heindel et al. 1988).

In the last few years, experimental data have been reported on the possible extension of the dissociation between explicit and implicit memory processes to individuals with intellectual disability (ID). Regarding repetition priming, studies investigating facilitation in identifying perceptually degraded pictures, because of previous exposure to the same pictures, have consistently reported a comparable priming effect in individuals with ID and in typically developing (TD) individuals matched for chronological age (CA) (Takegata & Furutuka 1993; Wyatt & Conners 1998) or mental age (MA) (Perrig & Perrig 1995). However, when repetition priming for verbal material was investigated, a complex, and somewhat contradictory, pattern of results emerged. Most of these studies were based on the Stem Completion procedure in which subjects are requested to complete a list of stems (i.e. the first three letters) with the first word that comes to mind. In this test, the priming effect is revealed by a bias in completing the stems with previously studied, rather than unstudied, words. Using this procedure, Carlesimo et al. (1997) and Vicari et al. (2000, 2001) reported a priming effect comparable with that of MA-matched TD children in various groups of individuals with ID [etiologically unspecified, Down syndrome (DS) and Williams syndrome (WS)]. However, using the same experimental paradigm, Mattson & Riley (1999) and Komatsu et al. (1996) reported reduced priming in groups of ID individuals with DS or unspecified aetiology.

Less experimental work has been devoted to investigating the ability of individuals with ID to learn visuo-motor or cognitive skills. In a first study, Vakil et al. (1997) compared the improvement in accuracy displayed by groups of individuals with ID and MA-matched children across successive trials of the Tower of Hanoi and Proteus Maze tests. On both tests, the individuals with ID performed significantly less accurately than the controls. However, while on the first test (which requires completing a spatial pattern according to a series of predetermined rules), the rate of trial-to-trial improvement was higher in the TD than in the ID group, in the Proteus Maze test (which requires solving a series of mazes with the least number of errors possible), the two groups improved at the same rate. Similar findings were reported more recently by Atwell et al. (2003) and by Vinter & Detable (2003).

Vicari et al. pointed out a difference in the skill-learning abilities of two genetically distinct groups of individuals with ID. In a first study (Vicari et al. 2000), a group of individuals with DS showed the same rate of improvement as a group of MA-matched TD children across successive trials of the Tower of London test (analogous to the Tower of Hanoi test) and in a comparison of the repeated vs. random blocks of a facilitated version of the serial reaction time (SRT) test originally developed by Nissen & Bullemer (1987). In this test, the subject was seated in front of a computer screen and a series of single coloured circles (green, blue and red) were presented centrally. The subject was asked to watch these stimuli and press the space bar as quickly as possible every time the green circle appeared on the screen. The test consisted of five blocks of 45 stimuli each. In the first block, the colour alternation was random and, therefore, unpredictable. In the next three sequences (II, III and IV), the colour alternation was strictly ordered (red, blue and green). In the final block, V, the order was again random. The software automatically recorded the time between the appearance of the stimulus on the screen and the subject's response (RT). If the subject learned the order in which the colours alternated on the screen, then the RT in the blocks formed by ordered sequences were gradually reduced with respect to the first random block, and, more importantly, worsened during the last random block.

In a second study, a group of children with WS showed significantly less procedural learning than TD children on both the Tower of London and SRT tests (Vicari et al. 2001). A subsequent study by Don et al. (2003) did not confirm reduced skill-learning capacities in individuals with WS. These authors compared a group of children and adults with WS and a group of TD individuals matched for CA in an artificial-grammar-learning paradigm (which requires to categorize letter-strings according to a specific rule) and in a rotor-pursuit task (where subjects are told to maintain contact between a stylus held in their preferred hand and a small disk on a rotating turntable). Although results documented evidence of skill learning in both groups, the rate of improvement in TD individuals was superior to that of the WS group in both tasks. However, the performance disadvantage in the WS group was no longer significant when group differences in working memory or non-verbal intelligence were taken into account.

That distinct etiological groups of individuals with ID perform differently on implicit memory tasks is relevant for the more general debate on the supposed asynchrony of cognitive maturation depending on the aetiology of the abnormal brain development. Indeed, following the developmental approach to ID (Zigler 1969; Zigler & Balla 1982), if individuals with and without ID of the same cognitive level (for instance, with similar MA) are compared, then a similar level of cognitive functioning is predicted in domains that involve learning, problem solving, information-processing and related psychological processes. In the same vein, children with ID of comparable cognitive level but different aetiology (i.e. distinct genetic syndromes) will exhibit similar neuropsychological profiles and cognitive abilities in ID individuals will develop uniformly irrespective of the specific aetiology of the abnormal brain maturation (‘syndrome independent’ theoretical perspective). According to this view, IQ and MA are the only predictors of ID individuals' performances on cognitive tasks, and no difference should be observed in the qualitative profile of cognitive impairment displayed by different etiological groups of ID, posed that they are comparable on general indexes of mental development.

In opposition to the developmental approach, other authors hold that cognitive maturation requires the development of a complex system of correlated, but also relatively independent, functions (Ellis 1969; Ellis & Cavalier 1982) and that in ID the prevalent impaired development of some functions rather than others may be at the base of the general intellectual deficit. Strictly related to this theoretical perspective, defined as the ‘difference approach’, is the ‘syndrome-specific’ position which claims that the development of cognitive abilities may be asynchronous among different etiological groups of ID individuals because of the different characteristics of their abnormal brain development. In this case, we should expect heterogeneous profiles of cognitive impairment even though individuals in the different etiological groups of ID are matched for general indexes of development such as IQ or MA.

This study was planned to verify the ‘etiological specificity’ hypotheses pertaining to the skill-learning abilities of individuals with ID. To this aim, we compared the performances of a group of individuals with DS and a group of individuals with WS, matched for CA and MA, on a modified version of the SRT task. If the hypothesis that skill learning is independent from aetiology in people with ID is correct, the two groups of ID individuals should present comparable profiles of learning. However, if the etiological specificity hypothesis is correct, DS and WS individuals should present heterogeneous profiles of skill learning, despite being similar for the overall level of cognitive dysfunction. In view of previous evidence collected in non-matched groups of DS and WS individuals (Vicari et al. 2000, 2001), we expected comparable skill-learning abilities in DS and matched TD participants but less than normal learning in WS individuals.

Methods

Results

Two of the 32 individuals with WS initially included in the study were excluded because they were unable to perform the SRT test. In accordance with what we reported in the method section, at the end of the sixth block, the individuals who performed the SRT were asked whether red square presentation was patterned or not. Moreover, all subjects were asked to reproduce the sequence on the keyboard. It emerged that four participants with WS and two TD children performed the SRT test using a declarative strategy; consequently, their performances were excluded from statistical analyses. Thus, all analyses were conducted on 26 participants with WS, 26 with DS and 47 TD children.

Procedural learning

Reaction times obtained by the three groups on the correct trials of the SRT test were analysed by means of a two-way mixed anova with Group as the between factor and Block (R1, O1, O2, O3, O4 and R2) as the within factor. The main effect of Group was significant, F_2,96= 10.3, P < 0.001, because the average RT of the participants with WS (M = 485.7, SD = 166.3) were faster than those of the TD children (M = 655.6, SD = 177.2; P < 0.001) and the DS individuals (M = 636.2, SD = 142.8; P < 0.01) who, instead, did not differ reciprocally (P > 0.10). The Block effect was also significant, F_5,480 = 52.5, P < 0.0001, thus demonstrating different RT in the whole group across blocks. In particular, RT progressively decreased passing from R1 to O1 block (P < 0.0001), decreasing further passing from O1 to O4 block (P < 0.0001), and then significantly increasing passing from O4 to R2 block (P < 0.0001). Crucial for the aims of this study, the Group–Block interaction was also significant, F_10,480 = 3.96, P < 0.0001, thus demonstrating a different pattern of RT changes in the three groups across blocks (Fig. 1). Planned comparisons made to qualify this interaction revealed a decline in RT passing from R1 to O1 block, which was significant in DS and TD groups (DS P < 0.001; TD P < 0.001) but not in WS; however, the between-group comparison for the rate of the RT decline did not reveal significant differences (WS: M = 76.4, SD = 156.0; DS: M = 130.6, SD = 165.0; TD: M = 137.9, SD = 187.6; P consistently >0.10). Similarly, the RT decrease passing from O1 to O4 block was also significant in DS and TD groups (DS P = 0.05; TD P < 0.001) but not in WS. In this case, the rate of decline was significantly larger in TD (M = 157.1, SD = 191.1) than in WS individuals (M = 62.3, SD = 120.8; P < 0.05), with no difference between DS participants (M = 105.9, SD = 136.2) and either TD or WS groups (P > 0.10 in both cases). Critical for the aims of the study was the significant group difference in the rebound effect passing from O4 to R2 block. In this case, in fact, the RT increase was significant in the TD and the DS individuals (P < 0.001 in both cases) but not in the WS participants. Indeed, the RT slowing exhibited by the WS group (M = 30.5, SD = 108.4) was significantly lower than that shown by both the TD (M = 192.0, SD = 150.2; P < 0.001) and the DS (M = 176.3, SD = 129.5; P = 0.001) groups which, instead, did not differ from each other.

An analogous, two-way Group–Block anova performed on the number of errors showed a significant Group effect, F_2,96 = 8.8, P < 0.001, because of the higher number of errors made by the individuals with WS (M = 31.7, SD = 21.1) than by TD (M = 16.5, SD = 12.7; P < 0.001) and the DS group (M = 16.0, SD = 15,9; P < 0.01). The effect of Block was significant, F_5,480 = 7.8, P < 0.0001, thus demonstrating a different degree of response accuracy in the whole group across blocks. Indeed, the number of errors progressively decreased passing from R1 to O1 block (P < 0.05) but was quite similar passing from O2 to O4 block (P > 0.10). As in the case of RT, the number of errors made by the whole group increased significantly passing from O4 to R2 block (P < 0.001). The Group–Block interaction was not significant, F_10,480 = 1.4, P > 0.10, indicating a similar variation in the number of errors in the three groups across blocks (Fig. 2).

Discussion

The two groups of individuals with ID who participated in the present study demonstrated different patterns of procedural learning. First, confirming some previous reports (Vicari et al. 2001, 2003), WS individuals revealed poor implicit learning of the temporal sequence of events characterizing the ordered blocks in the SRT task. Indeed, differently from normal controls, WS participants showed no RT speeding passing from O1 to O4 ordered block. Most importantly, the rebound effect, which so dramatically affected normal children's RT passing from O4 (ordered) to R2 (random) block, had only a marginal influence on WS children's RT. This demonstrates that while in the TD group, the RT reduction across the successive blocks of ordered sequences was largely mediated by learning the temporal sequence of events, in the WS group, there was no evidence of sequence learning. Note that the reduced implicit learning observed in people with WS cannot be explained by their basic difficulty in the processing of visual-spatial data or manual dexterity; indeed, no correlation was found between the learning index and performance scores on the SWM and constructive tasks.

Nonetheless, WS are quicker and make more mistakes than controls and, albeit the analyses of RT were performed on correct trials only, the possibility that accuracy-speed tradeoff could at least partially account for the group differences in performance must be considered.

Differently from the WS group, the rate of procedural learning of the participants with DS was comparable with that of their MA-matched controls. Indeed, DS and TD individuals showed parallel RT variations in the series of ordered blocks and, more importantly, passing from the O4 (ordered) to the R2 (random) block. Therefore, a substantial preservation of skill-learning abilities in this genetic syndrome is confirmed (Vicari et al. 2000; Krinsky-McHale et al. 2003).

Overall, the results of the present study document that procedural learning in individuals with ID depends on the aetiology of the syndrome. Thus, these results provide support for the etiological specificity account of cognitive development in individuals with ID. This conclusion is reinforced by the fact that differences in procedural learning between the DS and WS groups could not be accounted for by differences in general level of cognitive development. In fact, the two groups were comparable for both CA and MA; moreover, no significant association was found between an overall index of procedural learning and measures of cognitive development, such as IQ and MA, either in the two groups of ID individuals or in the group of TD children. The finding of IQ invariance in performances on implicit learning tasks is consistent with the literature on individuals without ID (Hasher & Zacks 1979; Reber et al. 1991) and individuals with mild ID (Ellis et al. 1989; Katz & Ellis 1991; Vicari et al. 2001, 2003; Krinsky-McHale et al. 2003, 2005).

The finding of non-homogeneous cognitive profiles in individuals with DS and WS is not new. Indeed, previous studies comparing individuals with DS and WS directly revealed discrepant performances in language (Mervis & Robinson 2000; Vicari et al. 2004), visual-spatial abilities (Wang & Bellugi 1993; Vicari et al. 2004) and memory functioning in both short- (Vicari & Carlesimo 2006; Brock 2007) and long-term domains (Vicari & Carlesimo 2002). The presently reported data extend to the implicit memory domain the evidence of qualitatively different patterns of cognitive functioning in WS and DS individuals.

Taken together, these data reinforce the view of ID as a variety of conditions in which some cognitive functions may be disrupted more than others (Vicari et al. 1992) and not as a unitary condition characterized by homogeneous slowness of cognitive development.

The different implicit memory profile exhibited by the various etiological groups of people with ID presumably results from some specific characteristics of their anomalous brain development. Recent neuroimaging studies (Galaburda & Bellugi 2000; Schmitt et al. 2001) have attempted to document the presence of particular morphological cerebral characteristics to explain the distinct cognitive and behavioural profiles observed in persons with ID, especially of known genetic syndromes. Concerning implicit memory, both neuropsychological (Molinari et al. 1997) and functional neuroimaging (Van Der Graaf et al. 2004) data assign a critical role to the basal ganglia and cerebellum in the implicit learning of visuo-motor skills. The brain development of individuals with WS is characterized by a remarkable atrophy of basal ganglia (Jernigan & Bellugi 1990), while the morphology of the cerebellum is generally normal (Wang et al. 1992). Further, a neurochemical alteration (reduction of the neurotransmitter N-acetylaspartate) has been reported (Rae et al. 1998) in the cerebellum of people with WS. The brains of individuals with DS, instead, exhibit severe cerebellar hypoplasia with normal morphology of the basal ganglia (Jernigan et al. 1993). In light of these data, we can tentatively conclude that the deficient maturation of visuo-motor skill learning in people with WS is related to the deficient maturation of the striatal circuits known to be critical for this ability.

In conclusion, we have described different implicit learning abilities in individuals with DS and WS of comparable MA and CA. Based on these data, we conclude that ID cannot be invariably associated with the preservation of implicit learning. Presumably, differences are related to biological differences in brain morphology and functionality determined by the different genetic alteration characteristic of each syndrome.