Implicit learning is defined as the acquisition of information or motor skill without conscious access to what was learned or even to the fact that learning occurred 
. Autism Spectrum Disorder (ASD) is characterized by social, communicative and motor impairments 
. The semantic and episodic memories of people with autism have often been studied, but neurocognitive studies of procedural learning and implicit cognition have received less attention. The extent of learning abilities of ASD individuals is debated 
. In the present study, we examined implicit motor skill learning in ASD to probe the functional integrity of this type of fundamental learning mechanism.
Most models of motor skill learning 
emphasize the role of the basal ganglia and the cerebellum, while the role of the hippocampus in this process remains inconclusive 
. Neuropsychological studies have shown that sequence learning is impaired in people with Huntington's and Parkinson's diseases 
, demonstrating the impact of striatal dysfunction on this type of perceptual-motor learning. Functional brain imaging studies also show the involvement of the cerebellum, striatum and motor cortices in implicit sequence learning tasks including the Serial Reaction Time (SRT) and the Alternating Serial Reaction Time (ASRT) tasks 
. In addition, Muller et al. 
reported that autistic individuals showed abnormal fMRI activity patterns in premotor cortex as well as greater individual variability in the activation maps.
Previous studies showed mixed results regarding implicit sequence learning of autistic people. Mostofsky and colleagues 
found impaired sequence learning when testing autistic children. They used the SRT task, developed by Nissen and Bullemer 
, in which participants were instructed to respond as quickly and as accurately as possible to the location of a stimulus that was presented at one of four possible locations on the monitor in a series of trials. Unknown to the participants, the locations of stimuli follow a predefined sequence, and participants typically become faster at responding to the locations predicted by the sequence compared to random trials. Mostofsky et al. 
tested a 10-trial fixed sequence repeated 8 times in a block, across a total of 5 blocks using a longer 1500 ms interval, instead of the customary 120–300 ms response-to-stimulus interval used in SRT tasks (e.g., 
). Gordon and Stark 
tested sequence learning in autistic participants in two tasks, one with an 8-element, and the other a 4-element fixed sequence. Their results revealed marginal learning with the 8-element fixed sequence task and significant learning with the 4-element task. As in Mostofsky et al. 
, this study used an unusually long response to stimulus interval (RSI) of 500 ms.
Four issues arise with the two studies above: 1) With a fixed-sequence series the possibility of an explicit strategy arises, because it is easier to become aware of the sequence, since the same sequence is presented repeatedly. 2) Both in the 10- and 8-element sequences the frequency of the elements was not balanced. Some elements could have occurred more frequently than others, which could increase the possibility of pattern recognition of the sequence, making the learning process explicit rather than implicit. In addition, it is possible that the learning observed was due at least partly to learning the relative frequencies of individual events rather than of sequences of events. 3) The long RSI values in the above studies could also contribute to developing an explicit strategy. Research has suggested that the longer the RSI, the more probable that explicit strategies are used 
. 4) In the various neuropsychological and neurodevelopmental disorders in which IQ is involved, it has been found that explicit learning is correlated with IQ, while implicit learning is relatively independent of IQ level 
. Explicit processes, therefore, suffer more under circumstances with IQ impairment. If learning relies on explicit strategies, then autistic individuals could be learning less than controls due to impairments in explicit rather than implicit learning.
Barnes and colleagues 
overcame the above limitations by using a 3-element version of the ASRT task 
, which is a modified version of the SRT task. In classical SRT tasks the structure of a sequence is deterministic with the stimuli following a simple repeating pattern as in the series 213412431423, where numbers refer to distinct events. In contrast, in the ASRT task 
repeating events alternate with random elements. This means that the location of every second stimulus on the screen is determined randomly. If, for instance, the sequence is 123, where the numbers represent locations on the screen, in ASRT the sequence of stimuli will be 1R2R3R1R2R3R…, with R representing a random element. The sequence is thus better hidden than in the classical SRT task and it is also possible to track sequence-specific learning continuously by comparing responses to the random and sequence elements within each testing block. This structure is called a probabilistic second-order (lag-2) dependency 
, because to predict element ‘n’ we need to know element n-2. Barnes et al. 
used a 120ms RSI, and they found intact learning in Autism compared to a control group matched for age and IQ. The authors suggest that the fronto-striatal-cerebellar functions are spared in autism.
It is possible that Barnes et al. 
found intact implicit learning because participants were mostly children with Asperger's syndrome, who have better cognitive abilities than children with simple autism. It is also possible that this group found intact implicit learning because they used the ASRT with 3 elements (i.e., 3 possible locations corresponding to 3 possible responses), which could be too easy to detect deficits. However, in a recent study Brown et al. 
also observed intact implicit sequence learning in a probabilistic SRT task introduced by Schvaneveldt & Gomez 
. In this task the RSI was 0 ms to reduce the possibility of creating an explicit strategy 
To our knowledge, consolidation of implicit or procedural learning has not been studied in autism, although some research has investigated consolidation of episodic and semantic long-term memories 
. Because some aspects of these domains show impairments in autism, it is important to investigate the implicit consolidation processes as well. When examining consolidation it is essential to know that skill learning occurs not only during practice in the so-called online
period, but also between-practice during the so- called offline
phase. The process that occurs during the offline period is referred to as consolidation, which means stabilization of a memory trace after the initial acquisition or even improvement in performance following an offline period 
. Such consolidation is important in considering the long-term acquisition of skills; even if implicit learning is intact, it is possible that autistic individuals are impaired in consolidation, thus forgetting the skills over the longer term. This might explain the apparent contradiction of intact implicit learning in autistic people even though they are known to be weaker in communicative and social skills 
In our study we used the ASRT task to investigate implicit learning and consolidation in autism. The ASRT task allows separation of general skill learning and sequence specific learning during both online and offline periods. General skill learning is reflected in the overall reaction time, whereas sequence-specific learning is reflected in the difference between the reaction time to predictable, sequence events as opposed to less predictable random ones. We also examined the effect of a 16-hour delay on learning performance, to test whether consolidation is intact. The present study goes beyond previous studies 
in two ways: 1) we used a more difficult 4-element ASRT task with 4 possible locations and 4 corresponding responses, instead of the 3-element version used by Barnes et al. 
, and 2) we investigated the consolidation of implicit learning over a 16-hour period.