内隐序列学习中转移组块的数量和位置效应

表征质量理论对意识增长持渐进观点,忽视新异刺激对意识的突变式影响;新异刺激理论强调意识突变,忽视新异刺激本身的表征质量增长。本研究采用经典确定性内隐序列学习范式,将转移组块作为新异刺激,操控其数量和位置,探究新异刺激如何通过表征质量来影响内隐学习和意识。结果表明：（1）数量效应显著,即两个转移组块更能促进内隐学习量,说明新异刺激本身需要足够的表征质量才能发挥“意外事件”的作用。（2）在位置效应上,两个转移组块且靠前的设置更能提高受控意识,表明第一个新异刺激必须出现在原序列初级表征质量阶段,才能使被试对新异刺激和原序列进行对比,加之第二个新异刺激的与之呼应,促进原序列意识增加。     

The memorial structure of organized sequences

Subjects were asked to memorize a sequence of nine consonants which were grouped into three groups of three letters each (e.g., SBJ FQL ZNG). After learning the sequence, they were presented with single letters, letter pairs, or letter triples and asked to indicate if the probe item appeared in the memorized sequence. The latency results suggest that subjects engage in a linear self-terminating memory search in which the items from one chunk are retrieved from memory as the items from the immediately preceding chunk are being scanned. When the probe consisted of more than one item, the subjects were slowed in their comparison if the letters came from different chunks (e.g., JF vs. FQ in the above illustration), and the number of letters in the probe also influenced the reaction time. Neither of those effects were obtained if all the letters in the probe came from the same chunk, and that would seem to suggest that the probe items from the same chunk were compared in parallel to the letters in the memory item, while items from different chunks were compared serial.     

Learning a keying sequence you never executed: Evidence for independent associative and motor chunk learning

A substantial amount of research has addressed how people learn and control movement sequences. Recent results suggested that practice with discrete key pressing sequences results in two types of sequence learning: associative learning and motor chunk development (Verwey & Abrahamse, 2012). In the present study, we addressed whether in keying sequences of limited length associative learning develops also when the use of the chunking mode is prevented by introducing during practice random deviants. In line with the notion of two different learning mechanisms, the present results indicate that associative sequence learning develops when motor chunks cannot be developed during practice. This confirms the notion that motor chunks do not rely on these associations. In addition, experience with a particular execution mode during the practice phase seems to benefit subsequent use of that mode with unfamiliar and random sequences. Also, participants with substantial video-gaming experience were faster in executing discrete keying sequences in the chunking mode. These last two results may point to the development of a general ability to produce movement sequences in the chunking mode.     

Kompositionsbildung bei Symbolfolgen und Bediensequenzen: Empirische Befunde und die Theorie des „Competitive Chunking”

In the paradigm of artificial grammar learning subjects have to reproduce regularly structured sequences of symbols. In a subsequent and unexpected test, they have to discriminate sequences that correspond to those of thememorization task from sequences that do not. Discriminability of grammatical and nongrammatical strings is judged as evidence for unconscious learning. This effect has been explained by the creation of chunks: Memorization of strings is facilitated by composing elements to chunks. The hierarchical organization of these units allows the representation of grammatical strings by a few high level chunks. The fewer chunks are necessary to represent a given string the higher is its familiarity and the probability to be judged grammatical. For the incidental acquisition of knowledge about operating a device the creation of chunks could also be shown. In contrast to letter series, two peculiarities were found: (a) discriminability was rather low, (b) irregularities at the end of input sequences were most effective. These peculiarities dissappeared, if — similar to letter strings — the information about a single operation of a sequence was permanently available. In the present simulations it was investigated whether these effects are consistent with the theory of competitive chunking. The temporal grain within sequences was varied to investigate the impact of the availability of sequential information on discriminability. Three different modes of chunk creation were simulated in the test phase to investigate whether the creation of chunks is obligatory for perceptual integration or requires a memorization task. The variation of temporal grain had only a marginal impact on discriminability. The simulation of the empirical data required that grammatical chunks are created within the test phase. Implications of these findings are discussed.     

Taking patterns for chunks: is there any evidence of chunk learning in continuous serial reaction-time tasks?

When exposed to a regular sequence, people learn to exploit its predictable structure. There have been two major ways of thinking about learning under these conditions: either as the acquisition of general statistical information about the transition probabilities displayed by the sequence or as a process of memorizing and using separate chunks that can later become progressively composed with extended practice. Even though chunk learning has been adopted by some theories of skill acquisition as their main building block, the evidence for chunk formation is scarce in some areas, and is especially so in the continuous serial reaction-time (SRT) task, which has become a major research tool in the study of implicit learning. This article presents a reappraisal, replication and extension of an experiment that stands so far as one of the few alleged demonstrations of chunk learning in the SRT task (Koch and Hoffmann, Psychological Res., 63:22–35, 2000). It shows that the effects which were taken as evidence for chunk learning can indeed be obtained before any systematic training and thus surely reflect a preexistent tendency rather than a learned outcome. Further analyses of the effects after extended practice confirm that this tendency remains essentially unchanged over continuous training unlike what could be expected from a chunk-based account of sequence learning.     

Visual statistical learning produces implicit and explicit knowledge about temporal order information and scene chunks: Evidence from direct and indirect measures

We examined whether visual statistical learning (VSL) produced implicit and/or explicit knowledge about temporal order information and scene chunks, using a rapid serial visual presentation target detection task and a two-alternative forced-choice (2AFC) familiarity test as indirect and direct measures of VSL, respectively. In the familiarization phase, participants viewed a visual stream of natural scenes consisting of four triplets (i.e., three images that always appeared in the same order). In the subsequent target detection task, participants were required to detect target items embedded in a stream of 12 images or 12 words representing each natural scene. In the final 2AFC familiarity test, participants observed two test sequences (statistically related triplets vs. unrelated foils) and decided whether the first or second sequence was more familiar based on the familiarization phase. In both test phases, we included the same (forward) and reverse (backward) order of scenes as presented during the familiarization phase to examine whether the expression of VSL was based on temporal order of scenes or scene chunks. The results of the target detection task showed a learning effect for both temporal order in the forward condition and chunks in the backward condition, irrespective of whether stimuli were images or words; in contrast, we did not observe a learning effect in the backward condition for scene images in the familiarity test. Our findings are compatible with a learning mechanism that has both implicit and explicit components based on temporal order information and scene chunks.     

The role of Gestalt grouping principles in visual statistical learning

A major issue in visual scene recognition involves the extraction of recurring chunks from a sequence of complex scenes. Previous studies have suggested that this kind of learning is accomplished according to Bayesian principles that constrain the types of extracted chunks. Here we show that perceptual grouping cues are also incorporated in this Bayesian model, providing additional evidence for the possible span of chunks. Experiment 1 replicates previous results showing that observers can learn three-element chunks without learning smaller, two-element chunks embedded within them. Experiment 2 shows that the very same embedded chunks are learned if they are grouped by perceptual cues, suggesting that perceptual grouping cues play an important role in chunk extraction from complex scenes.     

Chunking processes in the learning of event sequences: Electrophysiological indicators

The present study investigated whether effects of implicit learning (IL) are due to well-learned and explicitly represented parts of the stimulus material ("chunks"). To this purpose, event-related brain potentials (ERPs) were recorded during an oddball-version of a serial reaction time (RT) task: At unpredictable positions within a 16-item letter sequence, single deviant items replaced an item of the repeatedly presented standard sequence. After acquisition, the "process dissociation procedure" (Jacoby, 1991) was adopted to identify explicitly learned sequence parts for each participant. Acquisition of sequence knowledge was reflected in faster RTs for standard items than for deviant items and in enhanced N2b and P3b components for deviant items. While the ERP effects were obtained for explicitly represented sequence parts only, RT effects were independent of subsequent reproduction performance. These results indicated that (1) ERPs are a valid measure of explicit knowledge, (2) implicit and explicit knowledge coexist in serial RT tasks, and (3) chunking processes play a major role in the acquisition of explicit knowledge about event sequences.     

Operating characteristics of the implicit learning system supporting serial interception sequence learning

The memory system that supports implicit perceptual-motor sequence learning relies on brain regions that operate separately from the explicit, medial temporal lobe memory system. The implicit learning system therefore likely has distinct operating characteristics and information processing constraints. To attempt to identify the limits of the implicit sequence learning mechanism, participants performed the serial interception sequence learning (SISL) task with covertly embedded repeating sequences that were much longer than most previous studies: ranging from 30 to 60 (Experiment 1) and 60 to 90 (Experiment 2) items in length. Robust sequence-specific learning was observed for sequences up to 80 items in length, extending the known capacity of implicit sequence learning. In Experiment 3, 12-item repeating sequences were embedded among increasing amounts of irrelevant nonrepeating sequences (from 20 to 80% of training trials). Despite high levels of irrelevant trials, learning occurred across conditions. A comparison of learning rates across all three experiments found a surprising degree of constancy in the rate of learning regardless of sequence length or embedded noise. Sequence learning appears to be constant with the logarithm of the number of sequence repetitions practiced during training. The consistency in learning rate across experiments and conditions implies that the mechanisms supporting implicit sequence learning are not capacity-constrained by very long sequences nor adversely affected by high rates of irrelevant sequences during training.     

Evidence for lasting sequence segmentation in the discrete sequence-production task

It is well known that movement sequences are initiated and executed more slowly as they become longer. Those effects of sequence length, which have been found to lessen with practice, have been attributed to the development of a single motor chunk that represents the entire sequence. But an increasingly efficient distribution of programming can also explain the effects. To examine the mechanisms underlying skill in executing keying sequences, the authors examined the performance of participants (N = 18) who practiced a discrete sequence-production task involving fixed sequences of 3 and 6 key presses. Detailed examination of the effects of extensive practice, of regularities in key pressing order, and of a preceding choice RT task on the production of those sequences showed that most participants executed the 6-key sequence as 2 or more successive segments and continued to do so in the various conditions. The preceding choice RT task restored the sequence-length effect in latency that had disappeared with practice. The present results suggest that practice induces the development of motor chunks, each representing a short segment, and with longer sequences a control scheme for concatenating the motor chunks. Segmentation of longer sequences appeared to be concealed by individual segmentation differences unless there were regularities that imposed a common segmentation pattern.     

Can musical transformations be implicitly learned?

The dominant theory of what people can learn implicitly is that they learn chunks of adjacent elements in sequences. A type of musical grammar that goes beyond specifying allowable chunks is provided by serialist or 12-tone music. The rules constitute operations over variables and could not be appreciated as such by a system that can only chunk elements together. A series of studies investigated the extent to which people could implicitly (or explicitly) learn the structures of serialist music. We found that people who had no background in atonal music did not learn the structures, but highly selected participants with an interest in atonal music could implicitly learn to detect melodies instantiating the structures. The results have implications for both theorists of implicit learning and composers who may wish to know which structures they put into a piece of music can be appreciated.     

Ordering and reordering in the auditory and visual modalities

This study investigates the hypothesis that the modality effect (i.e., the often-found recall advantage of the last few items presented in the auditory- rather than visual-input modality) is attributable to a directional auditory trace. Such a directional trace should help performance of forward item-to-item recall, but should hurt performance if mental reordering is required. However, it was found in our Experiment 1 that mental item-to-item reordering of the materials did not affect the modality effect. In Experiment 2, strict control was exercised over the order of recall of chunks of items, as well as over the order of recall within chunks. It was found that the item-to-item reordering of the materials had little effect on the modality effect. However, the larger scale order of recall of the chunks had a large effect on the modality effect. If recall was in a forward order (i.e., first chunk, second chunk, third chunk), there was a dramatic superiority of the auditory materials on the last chunk. If the recall order was backward (third chunk, second chunk, first chunk), the modality effect was inconsequential. These results suggest that the modality effect is related to access to large memorial units rather than to item-to-item associations.     

What’s magic about magic numbers? Chunking and data compression in short-term memory

Short term memory is famously limited in capacity to Miller's (1956) magic number 7±2-or, in many more recent studies, about 4±1 "chunks" of information. But the definition of "chunk" in this context has never been clear, referring only to a set of items that are treated collectively as a single unit. We propose a new more quantitatively precise conception of chunk derived from the notion of Kolmogorov complexity and compressibility: a chunk is a unit in a maximally compressed code. We present a series of experiments in which we manipulated the compressibility of stimulus sequences by introducing sequential patterns of variable length. Our subjects' measured digit span (raw short term memory capacity) consistently depended on the length of the pattern after compression, that is, the number of distinct sequences it contained. The true limit appears to be about 3 or 4 distinct chunks, consistent with many modern studies, but also equivalent to about 7 uncompressed items of typical compressibility, consistent with Miller's famous magical number.     

Modelling unsupervised online-learning of artificial grammars: Linking implicit and statistical learning

Humans rapidly learn complex structures in various domains. Findings of above-chance performance of some untrained control groups in artificial grammar learning studies raise questions about the extent to which learning can occur in an untrained, unsupervised testing situation with both correct and incorrect structures. The plausibility of unsupervised online-learning effects was modelled with n-gram, chunking and simple recurrent network models. A novel evaluation framework was applied, which alternates forced binary grammaticality judgments and subsequent learning of the same stimulus. Our results indicate a strong online learning effect for n-gram and chunking models and a weaker effect for simple recurrent network models. Such findings suggest that online learning is a plausible effect of statistical chunk learning that is possible when ungrammatical sequences contain a large proportion of grammatical chunks. Such common effects of continuous statistical learning may underlie statistical and implicit learning paradigms and raise implications for study design and testing methodologies.     

Infants hierarchically organize memory representations

Throughout development, working memory is subject to capacity limits that severely constrain short‐term storage. However, adults can massively expand the total amount of remembered information by grouping items into chunks. Although infants also have been shown to chunk objects in memory, little is known regarding the limits of this ability. In particular, it remains unknown whether infants can create more complex memory hierarchies, binding representations of chunks into still larger chunks in recursive fashion. Here we tested the limits of early chunking, first measuring the number of items infants can bind into a single chunk and the number of chunks infants can maintain concurrently, and then, critically, whether infants can embed chunked representations into larger units. We tested 14‐month‐old infants' memory for hidden objects using a manual search task in which we manipulated memory load (the number of objects infants saw hidden) and the chunking cues provided. We found that infants are limited in the number of items they can chunk and in the number of chunks they can remember. However, we also found that infants can bind representations of chunks into ‘superchunks’. These results suggest that hierarchically organizing information strongly affects working memory, starting in infancy.     

Age-related differences in immediate serial recall: Dissociating chunk formation and capacity

We assessed the contribution of two hypothesized mechanisms to impaired memory performance of older adults in an immediate serial recall task: decreased temporary information storage in a capacity-limited mechanism, such as the focus of attention, and a deficit in binding together different components into cohesive chunks. Using a method in which paired associations between words were taught at varying levels to allow an identification of multiword chunks (Cowan, Chen, & Rouder, 2004), we found that older adults recalled considerably fewer chunks and, on average, smaller chunks than did young adults. Their performance was fairly well simulated by dividing attention in younger adults, unlike what has been found for long-term associative learning. Paired-associate knowledge may be used in an implicit manner in serial recall, given that younger adults under divided attention and older adults use it well despite the relatively small chunk capacities displayed by these groups.     

Position effects in chunked linear orders

Summary When someone is asked which of two items from a linearly ordered set of items comes first in that order, reaction time decreases with increasing distance between the two items in the given order. This is known as thedistance effect, a very robust finding in studies on linear orders. Surprisingly, investigations onchunked linear orders show unexpected diverse results. For example, reaction times todifferent probes (i. e., probes with items from different chunks) are sometimes longer than those tosame probes (i. e., probes with items from the same chunk). But if subjects' latency depends on the discriminability of chunk labels vs. exact position information within a chunk,different probes should always yield faster order judgments thansame probes. Two experiments investigated whether position effects, including both end-of-list and end-ofchunk effects, can account for the unexpected finding that between-chunk probes are sometimes answered more slowly than within-chunk probes. Moreover, the linea-rorder paradigm was extended to complex action sequences with more than two chunks, using original cooking recipes as materials. The results showed thatchunk-position effects play a crucial role in comparative judgments. Probes from the first and the last chunk were answered faster than those from a middle chunk.Border items (i. e., items next to a chunk boundary) did not produce a significant effect. However, chunking affected subjects' reaction time, as was indicated by the absence of a distance effect fordifferent probes. Besides, action sequences that are ordered in time, as well as concept lists used previously, can be considered linear orders, thus supporting the view that linear orders reflect a basic representation schema of human memory.     

Visual statistical learning in the newborn infant

Statistical learning – implicit learning of statistical regularities within sensory input – is a way of acquiring structure within continuous sensory environments. Statistics computation, initially shown to be involved in word segmentation, has been demonstrated to be a general mechanism that operates across domains, across time and space, and across species. Recently, statistical leaning has been reported to be present even at birth when newborns were tested with a speech stream. The aim of the present study was to extend this finding, by investigating whether newborns’ ability to extract statistics operates in multiple modalities, as found for older infants and adults. Using the habituation procedure, two experiments were carried out in which visual sequences were presented. Results demonstrate that statistical learning is a general mechanism that extracts statistics across domain since the onset of sensory experience. Intriguingly, present data reveal that newborn learner’s limited cognitive resources constrain the functioning of statistical learning, narrowing the range of what can be learned.     

Shortlist B: a Bayesian model of continuous speech recognition

A Bayesian model of continuous speech recognition is presented. It is based on Shortlist (D. Norris, 1994; D. Norris, J. M. McQueen, A. Cutler, & S. Butterfield, 1997) and shares many of its key assumptions: parallel competitive evaluation of multiple lexical hypotheses, phonologically abstract prelexical and lexical representations, a feedforward architecture with no online feedback, and a lexical segmentation algorithm based on the viability of chunks of the input as possible words. Shortlist B is radically different from its predecessor in two respects. First, whereas Shortlist was a connectionist model based on interactive-activation principles, Shortlist B is based on Bayesian principles. Second, the input to Shortlist B is no longer a sequence of discrete phonemes; it is a sequence of multiple phoneme probabilities over 3 time slices per segment, derived from the performance of listeners in a large-scale gating study. Simulations are presented showing that the model can account for key findings: data on the segmentation of continuous speech, word frequency effects, the effects of mispronunciations on word recognition, and evidence on lexical involvement in phonemic decision making. The success of Shortlist B suggests that listeners make optimal Bayesian decisions during spoken-word recognition.     

Emergence of rhythm during motor learning

Complex motor skill often consists of a fixed sequence of movements. Recent studies show that a stereotyped temporal pattern or rhythm emerges as we learn to perform a motor sequence. This is because the sequence is reorganized during learning as serial chunks of movements in both a sequence-specific and subject-specific manner. On the basis of human imaging studies we propose that the formation of chunk patterns is controlled by the cerebellum, its posterior and anterior lobes contributing, respectively, to the temporal patterns before and after chunk formation. The motor rhythm can assist the motor networks in the cerebral cortex to control automatic movements within chunks and the cognitive networks to control non-automatic movements between chunks, respectively. In this way, organized motor skill can be performed automatically and flexibly.