Translating visual information into action predictions: Statistical learning in action and nonaction contexts

Humans are sensitive to the statistical regularities in action sequences carried out by others. In the present eyetracking study, we investigated whether this sensitivity can support the prediction of upcoming actions when observing unfamiliar action sequences. In two between-subjects conditions, we examined whether observers would be more sensitive to statistical regularities in sequences performed by a human agent versus self-propelled ‘ghost’ events. Secondly, we investigated whether regularities are learned better when they are associated with contingent effects. Both implicit and explicit measures of learning were compared between agent and ghost conditions. Implicit learning was measured via predictive eye movements to upcoming actions or events, and explicit learning was measured via both uninstructed reproduction of the action sequences and verbal reports of the regularities. The findings revealed that participants, regardless of condition, readily learned the regularities and made correct predictive eye movements to upcoming events during online observation. However, different patterns of explicit-learning outcomes emerged following observation: Participants were most likely to re-create the sequence regularities and to verbally report them when they had observed an actor create a contingent effect. These results suggest that the shift from implicit predictions to explicit knowledge of what has been learned is facilitated when observers perceive another agent’s actions and when these actions cause effects. These findings are discussed with respect to the potential role of the motor system in modulating how statistical regularities are learned and used to modify behavior.     

Social intuition as a form of implicit learning: Sequences of body movements are learned less explicitly than letter sequences

In the current paper, we first evaluate the suitability of traditional serial reaction time (SRT) and artificial grammar learning (AGL) experiments for measuring implicit learning of social signals. We then report the results of a novel sequence learning task which combines aspects of the SRT and AGL paradigms to meet our suggested criteria for how implicit learning experiments can be adapted to increase their relevance to situations of social intuition. The sequences followed standard finite-state grammars. Sequence learning and consciousness of acquired knowledge were compared between 2 groups of 24 participants viewing either sequences of individually presented letters or sequences of body-posture pictures, which were described as series of yoga movements. Participants in both conditions showed above-chance classification accuracy, indicating that sequence learning had occurred in both stimulus conditions. This shows that sequence learning can still be found when learning procedures reflect the characteristics of social intuition. Rule awareness was measured using trial-by-trial evaluation of decision strategy (Dienes & Scott, 2005; Scott & Dienes, 2008). For letters, sequence classification was best on trials where participants reported responding on the basis of explicit rules or memory, indicating some explicit learning in this condition. For body-posture, classification was not above chance on these types of trial, but instead showed a trend to be best on those trials where participants reported that their responses were based on intuition, familiarity, or random choice, suggesting that learning was more implicit. Results therefore indicate that the use of traditional stimuli in research on sequence learning might underestimate the extent to which learning is implicit in domains such as social learning, contributing to ongoing debate about levels of conscious awareness in implicit learning.     

The influence of the response–stimulus interval on implicit and explicit learning of stimulus sequence

Three experiments investigated the influence of the response-stimulus interval (RSI) on implicit and explicit learning of stimulus sequences. Participants responded to numerals presented in predetermined positions with alternating long and short RSIs. Half of the participants were instructed explicitly to learn the position sequence. In the transfer phase of Experiments 1 and 2, changing RSI patterns reduced the expression of incidental and intentional learning of position sequence. In Experiment 3 the position sequence was transformed, except that sub-sequences demarcated by long RSIs remained unchanged; this greatly reduced the expression of intentional learning, and slightly reduced that of incidental learning. These results indicate that in implicit learning, stimulus sequences are learned under the constraints of RSIs, whereas in explicit learning, learning independent of RSIs, as well as learning constrained by RSIs, occurs.     

Explicit Awareness does not Modulate Retrograde Interference Effects in Sequence Learning

Motor sequences are learned explicitly or implicitly based on conscious awareness of the sequence. Interference happens when two sequences are learned successively. Here, we aimed to determine whether implicit and explicit sequence learning are affected differently by retrograde interference. Young healthy volunteers participated in either a control or interference group and either an explicit or implicit learning condition. We used a modified serial reaction time task to induce sequence learning and control awareness. Results showed that the overall amount of sequence learning was greater in the explicit condition compared to implicit. However, sequence learning was equally susceptible to retrograde interference under either condition. We conclude that although susceptible to interference, explicit awareness improves overall sequence learning compared to implicit conditions.     

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.     

Implicit versus explicit learning processes in a probabilistic, continuous fine-motor catching task

We compared the influences of explicit instruction and uninstructed implicit learning of correlations among visual events, in a fine-motor task. The task required visual tracking of a small "ball" of light and "catching" it by means of joystick manipulation. A general pattern of improvement with practice for instructed and "noninstructed" conditions was found. Additionally, both instructed and noninstructed conditions evidenced use of the predictive relationships among stimulus events. The improvement in overall performance caused by explicit verbal instructions was less than the improvement seen with the implicit learning condition. Further, instructed and noninstructed conditions showed different patterns of joystick activity. These findings suggest a qualitatively different approach to visual information processing and task performance for the instructional conditions     

Implicit visual learning: How the task set modulates learning by determining the stimulus–response binding

Implicit learning is one of the most fundamental learning mechanisms that enables humans to adapt to regularities inherent in the environment. Despite its high flexibility, it depends on constraints, such as selective attention. Here, we focused on the stimulus-to-response binding which defines the dimensions of the stimuli and the responses participants attend to. In a serial reaction time task with a visual sequence, we investigated whether this stimulus–response binding influences the amount of sequence learning. The results of Experiments 1 and 2 showed that visual sequence learning is reduced when participants do not attend to the relevant response dimension. Furthermore, the findings of Experiment 3 suggest that attention to the relevant response dimension increased the development of explicit knowledge without affecting implicit knowledge. This latter finding is difficult to reconcile with the assumption that explicit learning results from the gradual strengthening of sequence representations.     

Implicit learning of social predictions

Humans exchange a range of nonverbal social signals in every interaction. It is an open question whether people use these signals, consciously or unconsciously, to guide social behavior. This experiment directly tested whether participants could learn to predict another person’s behavior using nonverbal cues in a single interaction, and whether explicit knowledge of the cue-outcome relationship was necessary for successful prediction. Participants played a computerized game of rock-paper-scissors against an avatar they believed was another participant. Sometimes the avatar generated a predictive facial cue before the play. On these trials, participants’ win-frequency increased over time, even if they did not acquire explicit knowledge of the predictive cue. The degree to which participants could predict the avatar (wins on cued trials) related to their self-reported liking of the avatar. These findings demonstrate the importance of implicit associative learning mechanisms in guiding social behavior on a moment-to-moment basis during interaction.     

Evidence of automatic processing in sequence learning using process-dissociation

This paper proposes a way to apply process-dissociation to sequence learning in addition and extension to the approach used by Destrebecqz and Cleeremans (2001). Participants were trained on two sequences separated from each other by a short break. Following training, participants self-reported their knowledge of the sequences. A recognition test was then performed which required discrimination of two trained sequences, either under the instructions to call any sequence encountered in the experiment "old" (the inclusion condition), or only sequence fragments from one half of the experiment "old" (the exclusion condition). The recognition test elicited automatic and controlled process estimates using the process dissociation procedure, and suggested both processes were involved. Examining the underlying processes supporting performance may provide more information on the fundamental aspects of the implicit and explicit constructs than has been attainable through awareness testing.     

Perceived facial expressions of emotion as motivational incentives: evidence from a differential implicit learning paradigm

Participants (N = 216) were administered a differential implicit learning task during which they were trained and tested on 3 maximally distinct 2nd-order visuomotor sequences, with sequence color serving as discriminative stimulus. During training, 1 sequence each was followed by an emotional face, a neutral face, and no face, using backward masking. Emotion (joy, surprise, anger), face gender, and exposure duration (12 ms, 209 ms) were varied between participants; implicit motives were assessed with a picture-story exercise. For power-motivated individuals, low-dominance facial expressions enhanced and high-dominance expressions impaired learning. For affiliation-motivated individuals, learning was impaired in the context of hostile faces. These findings did not depend on explicit learning of fixed sequences or on awareness of sequence-face contingencies.     

Prediction in infants and adults: A pupillometry study

Adults use both bottom‐up sensory inputs and top‐down signals to generate predictions about future sensory inputs. Infants have also been shown to make predictions with simple stimuli and recent work has suggested top‐down processing is available early in infancy. However, it is unknown whether this indicates that top‐down prediction is an ability that is continuous across the lifespan or whether an infant's ability to predict is different from an adult's, qualitatively or quantitatively. We employed pupillometry to provide a direct comparison of prediction abilities across these disparate age groups. Pupil dilation response (PDR) was measured in 6‐month olds and adults as they completed an identical implicit learning task designed to help learn associations between sounds and pictures. We found significantly larger PDR for visual omission trials (i.e. trials that violated participants’ predictions without the presentation of new stimuli to control for bottom‐up signals) compared to visual present trials (i.e. trials that confirmed participants’ predictions) in both age groups. Furthermore, a computational learning model that is closely linked to prediction error (Rescorla‐Wagner model) demonstrated similar learning trajectories suggesting a continuity of predictive capacity and learning across the two age groups.     

Data-driven sequence learning or search: What are the prerequisites for the generation of explicit sequence knowledge?

In incidental sequence learning situations, there is often a number of participants who can report the task-inherent sequential regularity after training. Two kinds of mechanisms for the generation of this explicit knowledge have been proposed in the literature. First, a sequence representation may become explicit when its strength reaches a certain level (Cleeremans, 2006), and secondly, explicit knowledge may emerge as the result of a search process that is triggered by unexpected events that occur during task processing and require an explanation (the unexpected-event hypothesis; Haider & Frensch, 2009). Our study aimed at systematically exploring the contribution of both mechanisms to the generation of explicit sequence knowledge in an incidental learning situation. We varied the amount of specific sequence training and inserted unexpected events into a 6-choice serial reaction time task. Results support the unexpected-event view, as the generation of explicit sequence knowledge could not be predicted by the representation strength acquired through implicit sequence learning. Rather sequence detection turned out to be more likely when participants were shifted to the fixed repeating sequence after training than when practicing one and the same fixed sequence without interruption. The behavioral effects of representation strength appear to be related to the effectiveness of unexpected changes in performance as triggers of a controlled search.     

Performing the unexplainable: Implicit task performance reveals individually reliable sequence learning without explicit knowledge

Memory-impaired patients express intact implicit perceptual-motor sequence learning, but it has been difficult to obtain a similarly clear dissociation in healthy participants. When explicit memory is intact, participants acquire some explicit knowledge and performance improvements from implicit learning may be subtle. Therefore, it is difficult to determine whether performance exceeds what could be expected on the basis of the concomitant explicit knowledge. Using a challenging new sequence-learning task, robust implicit learning was found in healthy participants with virtually no associated explicit knowledge. Participants trained on a repeating sequence that was selected randomly from a set of five. On a performance test of all five sequences, performance was best on the trained sequence, and two-thirds of the participants exhibited individually reliable improvement (by chi-square analysis). Participants could not reliably indicate which sequence had been trained by either recognition or recall. Only by expressing their knowledge via performance were participants able to indicate which sequence they had learned.     

睡眠对知觉与动作序列内隐学习离线巩固效应的影响

离线阶段发生的学习被称为离线巩固, 即在最初获得知识之后, 即使没有额外的练习, 其记忆痕迹也会保持稳定或提高。有研究初步探究了睡眠对知觉和动作序列内隐学习离线巩固的影响, 然而, 这些研究未能实现知觉序列与动作序列的完全分离, 序列类型是否调节睡眠对内隐序列学习离线巩固的影响仍需进一步探讨。此外, 既往外显学习的研究发现相对于简单的序列, 复杂的序列更容易从睡眠中获益, 表现出基于睡眠的离线巩固效应。睡眠对知觉序列与动作序列内隐学习离线巩固的影响是否会受到序列复杂程度的调节尚不明确。为此, 本研究在完全分离知觉序列和动作序列的情况下, 通过3个实验操纵序列的长度及结构, 设置3种不同复杂程度的序列规则, 考察了这一问题。结果发现, 对于动作序列, 序列规则复杂程度较低时, 无论是否经过睡眠都会发生离线巩固效应, 而当动作序列规则较为复杂时, 只有经历睡眠才会引起离线巩固效应; 对于知觉序列, 无论何种难易程度的规则, 均未发生离线巩固效应。上述结果表明内隐序列知识基于睡眠的离线巩固会受到序列类型及序列复杂程度的调节, 这为内隐学习的离线巩固争论提供了新的视角。     

Sequence learning by action,observation and action observation

The serial reaction time (SRT) task was used to compare learning of a complex sequence by action (participants responded to sequential stimuli), by observation (participants watched but did not respond to sequential stimuli), and by action‐observation (participants watched an expert model responding to sequential stimuli). Each of these groups was compared with an untrained control group. Experiment 1 indicated that both observation and action‐observation were sufficient to support learning of a 12‐item second‐order conditional (SOC) sequence. Experiment 2 confirmed these findings, and showed that, as indexed by reaction time (RT), the extent of learning by observation and by action‐observation was comparable to that of action‐based learning. Using a recognition test, Experiment 2 and 3 also provided evidence that, whereas learning by stimulus observation was explicit, learning by action‐observation was implicit. These findings are consistent with a connection between motor systems and implicit learning, but do not support the hypothesis that overt action is necessary for implicit learning.     

The neural correlates of statistical learning in a word segmentation task: An fMRI study

Functional magnetic resonance imaging (fMRI) was used to assess neural activation as participants learned to segment continuous streams of speech containing syllable sequences varying in their transitional probabilities. Speech streams were presented in four runs, each followed by a behavioral test to measure the extent of learning over time. Behavioral performance indicated that participants could discriminate statistically coherent sequences (words) from less coherent sequences (partwords). Individual rates of learning, defined as the difference in ratings for words and partwords, were used as predictors of neural activation to ask which brain areas showed activity associated with these measures. Results showed significant activity in the pars opercularis and pars triangularis regions of the left inferior frontal gyrus (LIFG). The relationship between these findings and prior work on the neural basis of statistical learning is discussed, and parallels to the frontal/subcortical network involved in other forms of implicit sequence learning are considered.     

内隐序列学习与注意的关系

内隐序列学习与注意的关系是内隐学习研究领域中的一个热点问题。“双注意机制”理论、“抑制表达”理论和“干扰外显学习成分”理论,从不同的角度阐释了内隐序列学习与注意的关系,但支持证据和反对意见并存,论争激烈。该文试图在一定程度上整合这三种理论,提出了“动态需求”假设,认为内隐序列学习任务对注意的需求是动态变化的,随着内隐序列学习任务中外显学习成分比重的变大,其对注意资源的需求也会随之增加     

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.     

Evidence for separate representations for action and location in implicit motor sequencing

We examined sequential learning of actions in an experiment in which four different actions (push, twist, pinch, switch) were placed at four horizontal locations. At transfer, participants responded to a sequence that required performing the same sequence of actions at different locations and to a different sequence of actions at the same sequence of locations. Participants with explicit knowledge demonstrated only learning the sequence of response locations. However, participants with implicit knowledge learned the sequence of actions just as well as the sequence of locations, and performance on individual sequences was just as good as performance when both sequences were presented. These results demonstrated that two types of sequences, one of actions and another of response locations, can be learned simultaneously, suggesting that parallel representations are involved in implicit motor skill acquisition.