Representing word meaning and order information in a composite holographic lexicon

The authors present a computational model that builds a holographic lexicon representing both word meaning and word order from unsupervised experience with natural language. The model uses simple convolution and superposition mechanisms (cf. B. B. Murdock, 1982) to learn distributed holographic representations for words. The structure of the resulting lexicon can account for empirical data from classic experiments studying semantic typicality, categorization, priming, and semantic constraint in sentence completions. Furthermore, order information can be retrieved from the holographic representations, allowing the model to account for limited word transitions without the need for built-in transition rules. The model demonstrates that a broad range of psychological data can be accounted for directly from the structure of lexical representations learned in this way, without the need for complexity to be built into either the processing mechanisms or the representations. The holographic representations are an appropriate knowledge representation to be used by higher order models of language comprehension, relieving the complexity required at the higher level.     

An Attractor Model of Lexical Conceptual Processing: Simulating Semantic Priming

An attractor network was trained to compute from word form to semantic representations that were based on subject-generated features. The model was driven largely by higher-order semantic structure. The network simulated two recent experiments that employed items included in its training set (McRae and Boisvert, 1998). In Simulation 1, short stimulus onset asynchrony priming was demonstrated for semantically similar items. Simulation 2 reproduced subtle effects obtained by varying degree of similarity. Two predictions from the model were then tested on human subjects. In Simulation 3 and Experiment 1, the items from Simulation 1 were reversed, and both the network and subjects showed minimally different priming effects in the two directions. In Experiment 2, consistent with attractor networks but contrary to a key aspect of hierarchical spreading activation accounts priming was determined by featural similarity rather than shared superordinate category. It is concluded that semantic-similarity priming is due to featural overlap that is a natural consequence of distributed representations of word meaning.     

A probabilistic computational model of cross-situational word learning

Words are the essence of communication: They are the building blocks of any language. Learning the meaning of words is thus one of the most important aspects of language acquisition: Children must first learn words before they can combine them into complex utterances. Many theories have been developed to explain the impressive efficiency of young children in acquiring the vocabulary of their language, as well as the developmental patterns observed in the course of lexical acquisition. A major source of disagreement among the different theories is whether children are equipped with special mechanisms and biases for word learning, or their general cognitive abilities are adequate for the task. We present a novel computational model of early word learning to shed light on the mechanisms that might be at work in this process. The model learns word meanings as probabilistic associations between words and semantic elements, using an incremental and probabilistic learning mechanism, and drawing only on general cognitive abilities. The results presented here demonstrate that much about word meanings can be learned from naturally occurring child-directed utterances (paired with meaning representations), without using any special biases or constraints, and without any explicit developmental changes in the underlying learning mechanism. Furthermore, our model provides explanations for the occasionally contradictory child experimental data, and offers predictions for the behavior of young word learners in novel situations.     

Response trajectories reveal conflict phase in image–word mismatch

In the present study, response trajectories were used in a picture–word conflict task to determine the timing of intermediate processing stages that are relatively inaccessible to response time measures. A marker was placed above or below the word ABOVE or BELOW so that its location was congruent or in conflict with the word's meaning. To report either word location(above or below the marker) or word meaning, participants moved a mouse upward toward the appropriate top left or right answer corner on the display screen.Their response trajectories showed a number of distinctive features: First, at about 200 ms after stimulus onset(the "decision moment"), the trajectory abruptly began to arc toward the appropriate answer corner; second,when the word's meaning and position were in conflict,the trajectory showed an interruption that continued until the conflict was resolved. By varying the SOA of the word and marker onsets, we found that the word meaning and word position became available at approximately 325 ms and 251 ms, respectively, after their onsets, and that the delay to resolve conflicts was about 138 ms. The timing of these response trajectory events was more stable than any extracted from the final response times, demonstrating the power of response trajectories to reveal processing stages that are only poorly resolved, if at all, by response time measures [added].     

Word learning as Bayesian inference

The authors present a Bayesian framework for understanding how adults and children learn the meanings of words. The theory explains how learners can generalize meaningfully from just one or a few positive examples of a novel word's referents, by making rational inductive inferences that integrate prior knowledge about plausible word meanings with the statistical structure of the observed examples. The theory addresses shortcomings of the two best known approaches to modeling word learning, based on deductive hypothesis elimination and associative learning. Three experiments with adults and children test the Bayesian account's predictions in the context of learning words for object categories at multiple levels of a taxonomic hierarchy. Results provide strong support for the Bayesian account over competing accounts, in terms of both quantitative model fits and the ability to explain important qualitative phenomena. Several extensions of the basic theory are discussed, illustrating the broader potential for Bayesian models of word learning.     

The retention of nouns,adjectives, and verbs of sentences

This research compared the retention of nouns, adjectives and verbs of sentences. A model of the structure underlying the retention of sentences and in which deep structure constituent propositions are hierarchically related, was used to make predictions concerning the retention of these word classes. The model was not supported by the data and it was determined that latencies to recognize the words of a previously presented sentence were the same for all classes of words. This result indicated that for recognition (but not necessarily for other memory tasks), each word of a sentence is seperately or equally accessible (as long as inter-word imagery is the same). However, a further experiment showed that, although word class is not a variable, the image-arousing capacity of a word is. Latencies and error rates were much lower for high-imagery words than for low-imagery ones. The findings on error rates in recognizing words were interesting and pointed to a different model where the meaning representations and syntatic functions of words have an active role in sentence retention.     

The influence of word function in the missing-letter effect: Further evidence from French

When asked to detect target letters while reading continuous text, subjects miss more letters in highly common function words than in less common content words. This is known as themissing-letter effect. According to the structural account, the higher omission rates for frequent function words are attributable to their role in supporting the extraction of phrase structure, after which they become lost in the transition from structure to meaning. This implies that word function in and of itself should affect letter detection accuracy. This issue was examined in four experiments while controlling for a number of confounded factors associated with another influential model: the unitization account. The first experiment extended the missing-letter effect to the French language. The second showed that letter detection is influenced by slight variations in the function assumed by the same word, such as when it is used as a definite article as opposed to a pronoun. This effect was observed even when the frequency of the orthographic pattern and the syllable stress patterns were controlled. In the last two experiments, a control was added for another factor: frequency of word meaning. The results indicate that word function contributes to the missing-letter effect over and above what is contributed by frequency of word meaning.     

Contrasting effects of feature-based statistics on the categorisation and basic-level identification of visual objects

Conceptual representations are at the heart of our mental lives, involved in every aspect of cognitive functioning. Despite their centrality, a long-standing debate persists as to how the meanings of concepts are represented and processed. Many accounts agree that the meanings of concrete concepts are represented by their individual features, but disagree about the importance of different feature-based variables: some views stress the importance of the information carried by distinctive features in conceptual processing, others the features which are shared over many concepts, and still others the extent to which features co-occur. We suggest that previously disparate theoretical positions and experimental findings can be unified by an account which claims that task demands determine how concepts are processed in addition to the effects of feature distinctiveness and co-occurrence. We tested these predictions in a basic-level naming task which relies on distinctive feature information (Experiment 1) and a domain decision task which relies on shared feature information (Experiment 2). Both used large-scale regression designs with the same visual objects, and mixed-effects models incorporating participant, session, stimulus-related and feature statistic variables to model the performance. We found that concepts with relatively more distinctive and more highly correlated distinctive relative to shared features facilitated basic-level naming latencies, while concepts with relatively more shared and more highly correlated shared relative to distinctive features speeded domain decisions. These findings demonstrate that the feature statistics of distinctiveness (shared vs. distinctive) and correlational strength, as well as the task demands, determine how concept meaning is processed in the conceptual system.     

Semantic predictability eliminates the transposed-letter effect

Semantic predictability facilitates word recognition during language processing. One possible explanation for this facilitation is that highly specific predictions generated online during language processing preactivate some features of upcoming words. To explore whether, how, and when these predictions affect visual word recognition, in the two experiments reported here we investigated the influence of semantic predictability on transposed-letter priming. In order to do so, a paradigm that combines self-paced word-by-word reading with masked priming was developed. Transposed-letter priming occurred in nonconstraining contexts but not in constraining contexts, indicating that readers use context to make predictions about both letter identity and position in upcoming words, and that these predictions have an early influence on visual word recognition.     

Distinctive phonological features differ in relevance for both spoken and written word recognition

This paper discusses four experiments on Dutch which show that distinctive phonological features differ in their relevance for word recognition. The relevance of a feature for word recognition depends on its phonological stability, that is, the extent to which that feature is generally realized in accordance with its lexical specification in the relevant word position. If one feature value is uninformative, all values of that feature are less relevant for word recognition, with the least informative feature being the least relevant. Features differ in their relevance both in spoken and written word recognition, though the differences are more pronounced in auditory lexical decision than in self-paced reading.     

Modeling the effects of repetitions, similarity, and normative word frequency on old-new recognition and judgments of frequency

Judgments of frequency for targets (old items) and foils (similar; dissimilar) steadily increase as the number of times a target is studied increases, but discrimination of targets from similar foils does not steadily improve, a phenomenon termed registration without learning (D. L. Hintzman & T. Curran, 1995; D. L. Hintzman, T. Curran, & B. Oppy, 1992). The present experiment explores this phenomenon with words of differing normative word frequency. The retrieving-effectively-from-memory model (REM; R. M. Shifrrin & M. Steyvers, 1997, 1998) predicts that low-frequency words will be better recognized than high-frequency words because low-frequency words have more distinctive memory representations. A corollary of this assumption predicts that the typical recognition word-frequency effect will be disrupted when similar foils are tested. These predictions were confirmed, but to fit both the recognition and the judgment-of-frequency data, the authors used a "dual-process" extension of the REM model.     

Masked phonological priming effects in English: are they real? Do they matter?

For over 15 years, masked phonological priming effects have been offered as evidence that phonology plays a leading role in visual word recognition. The existence of these effects-along with their theoretical implications-has, however, been disputed. The authors present three sources of evidence relevant to an assessment of the existence and implications of these effects. First, they present an exhaustive meta-analytic literature review, in which they evaluate the strength of the evidence for masked phonological priming effects on English visual word processing. Second, they present two original experiments that demonstrate a small but significant masked priming effect on English visual lexical decision, which persists in conditions that may discourage phonological recoding. Finally, they assess the theory of visual word recognition offered by the DRC model (Coltheart, Rastle, Perry, Langdon, & Ziegler, 2001) in the context of their empirical data. Through numerous simulations with this model, they argue that masked phonological priming effects might best be captured by a weak phonological (i.e., dual-access) theory in which lexical decisions are made on the basis of phonological information.     

No ownership of common factors

Comments on the original article, "The efficacy of psychodynamic psychotherapy," by J. Shedler (see record 2010-02208-012). Shedler's informative article raised several issues worthy of comment. His choice of the word distinctive (p. 98) in describing aspects of psychodynamic technique is open to at least two interpretations. On the one hand, distinctive can have a qualitative meaning and indicate the presence of a characteristic that is not shared. For example, a sign in the Bronx Zoo distinguishes birds from all other creatures as follows: "If it has feathers it's a bird, if it doesn't, it isn't." On the other hand, distinctive can have a quantitative meaning and indicate that one practice has more of a common element than another practice. Careful reading of Shedler's article and the article by Blagys and Hilsenroth (2000) that forms the basis of the "seven features [that] reliably distinguished psychodynamic therapies from other therapies" (Shedler, 2010, p. 98) shows that Shedler subscribes to the latter, quantitative, definition of distinctive. In other words, the seven features he presented are present in both psychodynamic therapies and the cognitive-behavioral therapies to which he compares them. For example, although Shedler did not mention it, dialectical behavior therapy explicitly focuses on six of the seven features, namely, "focus on affect and expression of emotion," "exploration of attempts to avoid distressing thoughts and feelings," "identification of recurring themes and patterns," "discussion of past experience," "focus on interpersonal relations," and "focus on the therapy relationship" (Shedler, 2010, p. 99). However, in the articles that Blagys and Hilsenroth reviewed, psychodyamic therapists engaged in more of these behaviors than did cognitive-behavioral therapists.     

对汉字字形识别层次模型的实验验证

该研究采用Ｊｏｈｎｓｔｏｎ和ＭｃＣｌｅｌｌａｎｄ的实验范式以合体汉字为实验材料对层次模型进行了实验检验。结果表明,以假字掩蔽模式代替特征掩蔽模式使字目标对部件目标的字优效应明显下降。这一结果支持了层次模型,即合体汉字的字形加工以部件识别加工为中介。但本实验还发现了真字掩蔽模式下的字优效应量比假字掩蔽模式下明显减少,这一结果表明,在字加工水平中存在着觉察器间的相互竞争。因此,合体汉字字形识别的部件中介模型有待进一步完善。     

Interactions between study task, study time, and the low-frequency hit rate advantage in recognition memory

In studies of episodic recognition memory, low-frequency words (LF) have higher hit rates (HR) and lower false alarm rates (FAR) than do high-frequency words (HF), which is known as the mirror pattern. A few findings have suggested that requiring a task at study may reduce or eliminate the LF-HR advantage without altering the LF-FAR effect. Other studies have suggested that the size of the LF-HR advantage interacts with study time. To explore such findings more thoroughly and relate them to theory, the authors conducted 5 experiments, varying study time and study task. The full mirror pattern was found only in 2 cases: the standard condition requiring study for a later memory test and a condition requiring a judgment about unusual letters. The authors explain their findings in terms of the encoding of distinctive features and discuss the implications for current theories of recognition memory and the word frequency effect.     

The multidetermined nature of idiom processing

Models of idiom comprehension differ in their predictions concerning compositionality: Some claim that idiomatic meaning is the result of compositional analysis initiated at the earliest stages of comprehension, whereas others claim that compositional analysis occurs only at late stages, subsequent to direct retrieval--especially for idioms that are highly familiar. We evaluated these alternatives in four experiments by using a variety of online and offline comprehension measures. In Experiment 1, we analyzed the normative characteristics of 219 idioms with respect to these predictions. Dimensions of interest included several measures of decomposability, familiarity, and word frequency of the idioms' verbs and nouns. In Experiments 2 through 4, we determined how these dimensions relate to several online measures of idiom comprehension. High familiarity was associated with good comprehension across all experiments; however, facilitative effects of decomposability were found only for tasks that required an overt semantic judgment. Word frequency, but not semantic decomposability of the idiom-initial verb, was associated with comprehension for some measures. These data support a model of idiom comprehension, according to which figurative meaning arises from the time-dependent availability of multiple linguistic constraints, and in which decomposability plays a limited role in the earliest stages of idiom comprehension. Normative data for 210 of the idiomatic phrases may be downloaded from the Psychonomic Society Web archive at www.psychonomic.org/archive/.     

A word’s meaning affects the decision in lexical decision

The influence of an isolated word’s meaning on lexical decision reaction time (RT) was demonstrated through four experiments. Subjects in two experiments made lexical decision judgments, those in a third experiment pronounced the words used in the lexical decision task, and those in a fourth experiment quickly pronounced their first associative response to the words. Differences in lexical access time for the words were measured with the pronunciation task, and differences in meaning were assessed with the association task. Multiple regression analyses of lexical decision RT were conducted using associative RT, pronunciation RT, and other target word properties (printed frequency, length, instance dominance, and number of dictionary meanings) as predictor variables. These analyses revealed a relationship between lexical decision RT and associative RT after the effects of other variables had been partialed out. In addition, word frequency continued to have a significant relationship to lexical decision RT beyond that shared with pronunciation RT and the other variables. The results of these experiments indicate that at least some of the effect of word meaning and word frequency in lexical decision is attributable to a decision stage following lexical access.     

From concepts to lexical items.

In this paper we address the question how in language production conceptual structures are mapped onto lexical items. First we describe the lexical system in a fairly abstract way. Such a system consists of, among other things, a fixed set of basic lexical entries characterized by four groups of information: phonetic form, grammatical features, argument structure, and semantic form. A crucial assumption of the paper is that the meaning in a lexical entry has a complex internal structure composed of more primitive elements (decomposition). Some aspects of argument structure and semantic form and their interaction are discussed with respect to the issue of synonymy. We propose two different mappings involved in lexical access. One maps conceptual structures to semantic forms, and the other maps semantic forms to conceptual structures. Both mappings are context dependent and are many-to-many mappings. We present an elaboration of Levelt's (1989) model in which these processes interact with the grammatical encoder and the mental lexicon. Then we address the consequences of decomposition for processing models, especially the nature of the input of lexical access and the time course. Processing models that use the activation metaphor may have difficulties accounting for certain phenomena where a certain lemma triggers not one, but two or more word forms that have to be produced with other word forms in between.     

A hybrid-similarity exemplar model for predicting distinctiveness effects in perceptual old-new recognition

In 2 sets of experiments, the authors investigated the basis for old-item distinctiveness effects in perceptual recognition, whereby distinctive old items are recognized with higher probability than are typical old items. In Experiment 1, distinctive old items were defined as those lying in isolated regions of a continuous-dimension similarity space. In this case, any beneficial effects of distinctiveness were absent or small, regardless of the structure of the test list used to assess recognition memory. In Experiment 2, distinctive items were defined as those objects containing certain discrete, individuating features. In this case, large old-item distinctive effects were observed, with the nature of the effects being modulated by the structure of the test lists. A hybrid-similarity exemplar model, combining elements of continuous-dimension distance and discrete-feature matching, was used to account for these distinctiveness effects in the recognition data.     

Lexical Acquisition in Probable Alzheimer''s Disease

Patients with probable Alzheimer's disease (pAD) were exposed to a new verb in a naturalistic fashion. We probed their knowledge of the word's semantic and grammatical characteristics for several minutes following this exposure, and compared this with their performance on parallel measures assessing known words. Significant differences were seen between pAD patients and controls in the acquisition of the new verb's semantic meaning and its argument structure, but pAD patients did not differ from controls in the acquisition of the new word's grammatical form class. Individual patient analyses demonstrated parallel deficits restricted to the semantic meaning and argument structure of the new word and known words in several pAD patients, suggesting that a selective language impairment contributed to their word learning deficit. This pattern is consistent with an intimate relationship between semantic meaning and argument structure in semantic memory. Other pAD patients had difficulty learning about all aspects of the new word, despite good performance with known words, suggesting that compromised memory may have limited their lexical acquisition.