Learning and development in neural networks--the importance of prior experience

Infants can discriminate between familiar and unfamiliar grammatical patterns expressed in a vocabulary that is distinct from that used earlier during familiarization (Cognition 70(2) (1999) 109; Science 283 (1999) 77). Various models have captured the data, although each required that discrimination be distinct, in terms of the computational process, from familiarization. This article describes a simple recurrent network (SRN), equipped only with the assumption that it should predict what comes next, which models the data without distinguishing between familiarization and discrimination. To accomplish this, the SRN requires pre-training on a range of sequences instantiating different structures and different vocabulary items to those used subsequently during familiarization and test. Pre-training enables the network to avoid replacing structure acquired during familiarization with structure experienced at test. An equivalent enabling condition may underpin infants' resistance to catastrophic interference between the different structures and vocabulary items to which they are exposed.     

Treating a Chronically Depressed Adolescent Female Using the Cognitive Behavioral Analysis System of Psychotherapy: A Case Study

A case is presented of an adolescent female with double depression who was treated using the Cognitive Behavioral Analysis System of Psychotherapy (CBASP). CBASP is designed to teach a social problem-solving procedure called Situational Analysis (SA). Generalized treatment effects were measured through monitoring of diagnostic status, two administrations of the Minnesota Multiphasic Personality Inventory-Adolescent, weekly administrations of the Child Depression Inventory, and acquisition performance ratings on the SA task. The results showed that CBASP appeared to be an effective treatment. The double depressive disorder remitted, and the patient learned to identify behavioral consequences as well as target and attain her interpersonal goals. CBASP can be effective for adolescents, although modifications of the adult form of the therapy may be necessary.     

Development of rules and strategies: balancing the old and the new

The experiments described in the lead articles replicate findings from previous studies of development of knowledge about balance scales, add several new findings, and raise four key questions: (a) How can rule use best be assessed? (b) How can we reconcile systematic use of rules with variable use of strategies? (c) When do children begin to use rules? and (d) How do children generate new rules? In this Reflection, we summarize current understanding of development of knowledge about balance scales and consider each of the four questions.     

A new approach to the formation of equivalence classes in pigeons

Four pigeons were given simultaneous discrimination training with visual patterns arbitrarily divided into two sets, with the stimuli in one set designated A1, B1, C1, and D1 and those in the other set designated A2, B2, C2, and D2. In sequentially introduced training phases, the pigeons were exposed to a series of reversals to establish AB and then CD equivalences. In subsequent testing sessions, a subset of stimuli from one set served as positive stimuli and those from the other set as negative stimuli on training trials, and transfer of the reinforced relation to other members of the sets was tested with nonreinforced probe trials. The pigeons were trained further on AC and BD equivalences and then were tested for the emergence of untrained AD and BC equivalences. Two of the 4 pigeons exhibited the emergence of one of these untrained equivalences, evidence for the emergence of transitive relations. This finding suggests that the pigeons established three-member functional equivalence classes by incorporating separately trained multiple equivalence relations. Repeated reversal training and probe testing enabled us to explore the formation and expansion of functional equivalence classes in pigeons.     

A multisource exploration of the friendship patterns of children with and without learning disabilities

Friendship patterns of 117 children with learning disabilities (LD) and 115 children without LD in Grades 4–8 were examined. In comparison with children without LD, boys with LD had fewer mutual friends, children with LD had more friends with learning problems and more younger friends, and children with LD in Grades 4–6 had less stable relationships. With regard to friendship quality, children with LD reported higher levels of conflict, lower levels of validation, and more problems with relationship repair than did children without LD. The findings were discussed in terms of factors that have been found to enhance friendship such as proximity and similarity, and the social skills difficulties that have been associated with learning disabilities.     

小学儿童对日常生活事件时间关系推理能力的初探

探查小学儿童对含有“在什么之前”“同时”日常生活事件时间关系的推理发展过程及其影响因素。被试为7岁、9岁、11岁3个年龄组共54名小学儿童,每个年龄组18人。4种问题模式分别为带有传递关系的单模型问题、带有无关前提的单模型问题、含两个前提的单模型问题、有肯定答案的多模型问题。每种问题模式含3种具体探察任务。主要研究结果表明：在本实验条件下,（1）儿童对事件时间顺序关系的推理从7岁到9岁有快速发展的趋势;（2）单模型和多模型任务无显著差异,模型数量似乎不是影响儿童推理成绩的关键因素;（3）儿童能够主动采用5种策略解决问题。逆向关系传递策略是问题解决的有效策略。     

Rats use a sense of direction to alternate on T-mazes located in adjacent rooms

Lister hooded rats were trained on a forced-sample T-maze alternation task in an environment lacking spatial landmarks. An early study of spontaneous alternation on the T-maze had shown that rats use a "spatial sense" to select alternate maze arms across mazes. As this phenomenon may provide a useful tool for studying the neural substrates of a directional sense, we wished to confirm this finding on a different version of the T-maze task, with well-trained animals. We found that rats successfully selected the appropriate maze arm when the choice phase of the task was presented on a second maze, oriented in the same direction, and located in an adjacent room. However, choice performance fell to chance level when the second maze was oriented 90° relative to the first. This result suggests that the rats do not simply alternate turns across the two environments, but rather that they rely on a sense of direction that is carried across environments. Electronic Publication     

One Frog, Two Frog, Red Frog, Blue Frog: Factors Affecting Children's Syntactic Choices in Production and Comprehension

Two experiments are reported which examine children's ability to use referential context when making syntactic choices in language production and comprehension. In a recent on-line study of auditory comprehension, Trueswell, Sekerina, Hill, and Logrip (1999) examined children's and adults' abilities to resolve temporary syntactic ambiguities involving prepositional phrases (e.g., “Put the frog on the napkin into ¨”). Although adults and older children used the referential context to guide their initial analysis (pursuing a destination interpretation in a one-frog context and a modifier interpretation in a two-frog context), 4 to 5-year olds' initial and ultimate analysis was one of destination, regardless of context. The present studies examined whether these differences were attributable to the comprehension process itself or to other sources, such as possible differences in how children perceive the scene and referential situation. In both experiments, children were given a language generation task designed to elicit and test children's ability to refer to a member of a set through restrictive modification. This task was immediately followed by the “put” comprehension task. The findings showed that, in response to a question about a member of a set (e.g., “Which frog went to Mrs. Squid's house?”), 4- to 5-year-olds frequently produced a definite NP with a restrictive prepositional modifier (e.g., “The one on the napkin”). These same children, however, continued to misanalyze put instructions, showing a strong avoidance of restrictive modification during comprehension. Experiment 2 showed that an increase in the salience of the platforms that distinguished the two referents increased overall performance, but still showed the strong asymmetry between production and comprehension. Eye movements were also recorded in Experiment 2, revealing on-line parsing patterns similar to Trueswell et al.: an initial preference for a destination analysis and a failure to revise early referential commitments. These experiments indicate that child–adult differences in parsing preferences arise, in part, from developmental changes in the comprehension process itself and not from a general insensitivity to referential properties of the scene. The findings are consistent with a probabilistic model for uncovering the structure of the input during comprehension, in which more reliable linguistic and discourse-related cues are learned first, followed by a gradually developing ability to take into account other more uncertain (or more difficult to learn) cues to structure.     

The Client Is the Most Important Common Factor: Clients' Self-Healing Capacities and Psychotherapy

I first briefly review the dodo bird verdict and suggest that we should be responding to it by looking for a new way to conceptualize how therapy works. Then I describe the dominant medical or treatment model of psychotherapy and how it puts the client in the position of a dependent variable who is operated on by supposedly potent therapeutic techniques. Next I argue that the data do not fit with this model. An alternative model is that the client is the most important common factor and that it is clients' self-healing capacities which make therapy work. I then argue that therapy has two phases—the involvement phase and the learning phase—and that the involvement phase is the most important. I next review the five learning opportunities provided by therapy. Finally, I argue that a relational model of therapy focused on consultation, collaboration, and dialogue is better than a treatment model.     

The variability of practice hypothesis in motor learning: does it apply to Alzheimer's disease?

Based on Schmidt's (1975) variability of practice hypothesis, this study examined acquisition and transfer of a gross motor skill, namely tossing, in 58 patients with Alzheimer's disease (AD) and 58 healthy older adults under constant, blocked, and random practice conditions. While healthy older adults were able to learn the tossing task equally well under the three practice conditions, only AD patients receiving constant practice showed significant improvements. Tests of intermediate transfer yielded the expected random practice advantage in healthy controls but not AD patients. None of the practice conditions facilitated intermediate transfer in AD patients; however, constant practice did benefit these impaired individuals on tests of near transfer. These results indicate that the variability of practice hypothesis does not extend to AD patients. As motor learning and transfer were clearly a function of constant practice, future attempts to retrain basic activities of daily living in AD patients should emphasize consistency in training.