Nested incremental modeling in the development of computational theories: the CDP+ model of reading aloud

At least 3 different types of computational model have been shown to account for various facets of both normal and impaired single word reading: (a) the connectionist triangle model, (b) the dual-route cascaded model, and (c) the connectionist dual process model. Major strengths and weaknesses of these models are identified. In the spirit of nested incremental modeling, a new connectionist dual process model (the CDP+ model) is presented. This model builds on the strengths of 2 of the previous models while eliminating their weaknesses. Contrary to the dual-route cascaded model, CDP+ is able to learn and produce graded consistency effects. Contrary to the triangle and the connectionist dual process models, CDP+ accounts for serial effects and has more accurate nonword reading performance. CDP+ also beats all previous models by an order of magnitude when predicting individual item-level variance on large databases. Thus, the authors show that building on existing theories by combining the best features of previous models--a nested modeling strategy that is commonly used in other areas of science but often neglected in psychology--results in better and more powerful computational models.     

How predictable is spelling? Developing and testing metrics of phoneme-grapheme contingency

Prior research has purported to show that words with infrequent phoneme-grapheme correspondences are more difficult to spell than words with frequent phoneme-grapheme correspondences. Defining exactly what a phoneme-grapheme relationship is, however, is not necessarily straightforward. There are a number of different assumptions that can be made. In this study, we developed four metrics of sound-spelling contingency based on all monosyllabic English words, including those with complex morphology. These metrics differed in the extent to which they included assumptions about spelling. The psychological reality of these different metrics was evaluated against data from a large-scale study of skilled adult spelling and a nonword spelling experiment. The results suggest that when spelling, people are sensitive to positional information, morphological status, vowel type, and a number of more idiosyncratic constraints.     

Visual phonology: The effects of orthographic consistency on different auditory word recognition tasks

In this study, we investigated orthographic influences on spoken word recognition. The degree of spelling inconsistency was manipulated while rime phonology was held constant. Inconsistent words with subdominant spellings were processed more slowly than inconsistent words with dominant spellings. This graded consistency effect was obtained in three experiments. However, the effect was strongest in lexical decision, intermediate in rime detection, and weakest in auditory naming. We conclude that (1) orthographic consistency effects are not artifacts of phonological, phonetic, or phonotactic properties of the stimulus material; (2) orthographic effects can be found even when the error rate is extremely low, which rules out the possibility that they result from strategies used to reduce task difficulty; and (3) orthographic effects are not restricted to lexical decision. However, they are stronger in lexical decision than in other tasks. Overall, the study shows that learning about orthography alters the way we process spoken language.     

Changes in psychological performance parameters of concentration and critical flicker-fusion frequency with advancing age and in relation to diffuse cerebral arterial affections of the circulation

Psychometric performance parameters and the critical flicker-fusion frequency are studied in various groups and in patients with chronic cerebral circulation trouble. Test d2 proves to be effective in the measuring of such impaired performance. There is a higher degree of correlation with Paul's addition test. A correlation between the performance parameters and the patient's age can be statistically substantiated. Among patients up to one third in the age group 55 and over show and affection of the cerebral performance turbances in spite of the absence of pathological clinical findings. The figure is twice as high in the group of patients undergoing treatment. The increased occurrence in old age shows the significance of pathological processes. There appears to be no relationship between critical flicker-fusion frequency and performance and stress factors. Individual correlations would support other physiological relations, such as an at present only hypothetical link with the psychic state (nervousness) of the test person. Our tables are suitable for practical purposes.     

MacLaboratory Controller: A switch and A/D interface between Apple’s Macintosh and peripheral apparatus

Apple’s Macintosh microcomputer has advantages over other systems used for laboratory control and data acquisition, especially because of its graphic and user-friendly features. However, a major perceived limitation has been its closed architecture. We describe an interface using the RS 422 modem port that supports up to 64 input/output (I/O) functions including analog-to-digital 4–12 channel input. The Controller software supplies an easily edited authoring tool in which functional relationships, logical operators, timing (millisecond accuracy), and counting functions are established through simple mouse and menu commands. Data from I/O functions may be displayed or saved as tab-delimited files that can be opened by various Macintosh statistical and spreadsheet packages and that are transportable to some mainframe applications.     

Visual search for singleton feature targets within and across feature dimensions

Three experiments investigated visual search for singleton feature targets. The critical dimension on which the target differed from the nontargets was either known in advance or unknown—that is, the critical difference varied eitherwithin a dimension oracross dimensions. Previous work (Treisman, 1988) had shown that, while the search reaction time (RT) functions were flat in both conditions, there was an intercept cost for the cross-dimension condition. Experiment 1 examined whether this cost would disappear when responses could be based on the detection ofany (target—nontarget) difference in the display (by requiring a “heterogeneity/homogeneity” decision). The cost remained. This argues that pop-out requires (or involves) knowledge of the particular dimension in which an odd-one-out target differs from the nontargets; furthermore, that knowledge is acquired through the elimination of dimensions not containing a target. In Experiment 2, the subjects had to eliminate (or ignore) one potential source of difference in order to give a positive response (displays could contain a “noncritical” difference requiring a negative response). The result was a comparatively large cost in the within-dimension (positive) condition. This can be taken to indicate that popout as such does not make available information as to the particular feature value in which the target differs from the nontargets. Experiment 3 examined whether search priorities can be biased in accordance with advance knowledge of the likely source of difference. The subjects were found to have a high degree of top-down control over what particular dimension to assign priority of checking to. The implication of the results for models of visual search and selection are discussed.     

Beyond rhyme or reason: ERPs reveal task-specific activation of orthography on spoken language

Metaphonological tasks, such as rhyme judgment, have been the primary tool for the investigation of the effects of orthographic knowledge on spoken language. However, it has been recently argued that the orthography effect in rhyme judgment does not reflect the automatic activation of orthographic codes but rather stems from sophisticated response strategies. Such a claim stands in sharp contrast with recent findings using event-related brain potentials (ERPs) in lexical and semantic tasks, which were taken to suggest that orthographic information occurs early enough to affect the core process of lexical access. Here, we show that the electrophysiological signature of the orthography effect in rhyme judgment is indeed different from the one obtained in online lexical or semantic tasks. That is, we did not find the orthography effect in the 300-350 ms time window which has previously been shown to process lexical information in the lexico-semantic tasks, but the effect appeared within the 175-250 ms and the 375-750 ms time-windows which we interpreted to reflect segmentation and decisional process, respectively. We conclude that the interactions between phonology and orthography are task-specific. Metaphonological tasks appear of limited use for studying the core processes and interactions that underlie lexical access.     

Relationship between patient characteristics and treatment allocation for patients with personality disorders

Within a large multi-center study in patients with personality disorders, we investigated the relationship between patient characteristics and treatment allocation. Personality pathology, symptom distress, treatment history, motivational factors, and sociodemographics were measured at intake in 923 patients, who subsequently enrolled in short-term or long-term outpatient, day hospital, or inpatient psychotherapy for personality pathology. Logistic regressions were used to examine the predictors of allocation decisions. We found a moderate relationship (R(2) = 0.36) between patient characteristics and treatment setting, and a weak relationship (R(2) = 0.18) between patient characteristics and treatment duration. The most prominent predictors for setting were: symptom distress, cluster C personality pathology, level of identity integration, treatment history, motivation, and parental responsibility. For duration the most prominent predictor was age. We conclude from this study that, in addition to pathology and motivation factors, sociodemographics and treatment history are related to treatment allocation in clinical practice.     

Do current connectionist learning models account for reading development in different languages?

Learning to read a relatively irregular orthography, such as English, is harder and takes longer than learning to read a relatively regular orthography, such as German. At the end of grade 1, the difference in reading performance on a simple set of words and nonwords is quite dramatic. Whereas children using regular orthographies are already close to ceiling, English children read only about 40% of the words and nonwords correctly. It takes almost 4 years for English children to come close to the reading level of their German peers. In the present study, we investigated to what extent recent connectionist learning models are capable of simulating this cross-language learning rate effect as measured by nonword decoding accuracy. We implemented German and English versions of two major connectionist reading models, Plaut et al.'s (Plaut, D. C., McClelland, J. L., Seidenberg, M. S., & Patterson, K. (1996). Understanding normal and impaired word reading: computational principles in quasi-regular domains. Psychological Review, 103, 56-115) parallel distributed model and Zorzi et al.'s (Zorzi, M., Houghton, G., & Butterworth, B. (1998a). Two routes or one in reading aloud? A connectionist dual-process model. Journal of Experimental Psychology: Human Perception and Performance, 24, 1131-1161); two-layer associative network. While both models predicted an overall advantage for the more regular orthography (i.e. German over English), they failed to predict that the difference between children learning to read regular versus irregular orthographies is larger earlier on. Further investigations showed that the two-layer network could be brought to simulate the cross-language learning rate effect when cross-language differences in teaching methods (phonics versus whole-word approach) were taken into account. The present work thus shows that in order to adequately capture the pattern of reading acquisition displayed by children, current connectionist models must not only be sensitive to the statistical structure of spelling-to-sound relations but also to the way reading is taught in different countries.     

Effects of phonological and orthographic neighbourhood density interact in visual word recognition

The present study investigated the role of phonological and orthographic neighbourhood density in visual word recognition. Three mechanisms were identified that predict distinct facilitatory or inhibitory effects of each variable. The lexical competition account predicts overall inhibitory effects of neighbourhood density. The global activation (familiarity) account predicts overall facilitatory effects of neighbourhood density. Finally, the cross-code consistency account predicts an interaction, with inhibition of phonological neighbours in sparse orthographic regions and facilitation of phonological neighbours in dense orthographic regions. In Experiment 1 (lexical decision), a cross-over interaction was indeed found, supporting the prediction of the cross-code consistency account. In Experiment 2, this cross-over interaction was exaggerated by adding pseudohomo-phone stimuli (e.g., brane) among the nonword targets. Finally, in Experiment 3 (progressive demasking), we tried to shift the balance between inhibitory and facilitatory mechanisms by using a perceptual identification task. As predicted, the inhibitory effects of phonological neighbourhood were amplified, whereas the facilitatory effects disappeared. We conclude that the level of compatibility across co-activated orthographic and phonological representations is a major causal factor underlying this pattern of effects.