You are in: Home » History of the Unit
2. COGNITIVE PSYCHOLOGY
General Notes. This material has been scanned from the original typescript. While we have done our best to remove errors, some may well remain. You can access other parts of this particular Progress Report either from the menu at the bottom of this entry, by using your browser's back function, by navigating back to the Unit History Timeline, or by accessing the relevant section of the electronic archive. Reference for this report are indexed by number and these can be found in a dedicated section also accessible from the menu at the bottom of this entry.
2.1 STUDIES OF LONG- AND SHORT-TERM MEMORY (Project 32) (Baddeley, Edworthy, Eldridge, Gathercole, Idzikowski, V.Lewis, Logie, Nimmo-Smith)
2.1.1 Working memory
Working memory refers to the system that underlies the temporary storage of information that is necessary for performing cognitive tasks such as thinking, reading and learning. A model of working memory that assumes a limited capacity controller, the Central Executive, aided by two active slave systems, the Articulatory Loop for verbal material and the Visuo-spatial Sketchpad for spatial material has continued to prove useful, and a decade of work in the area has been reviewed both relatively briefly (8) and in a forthcoming book (18 U). While a number of studies have been directly aimed at developing the underlying model (22; 20), the essence of the concept of working memory is its concern with the application of memory to other cognitive tasks. Our work has therefore concentrated primarily on studying memory through other cognitive skills and their breakdown, a feature-which makes the assignment of research in this area to a particular project somewhat arbitrary. They are reported together since they do have a common central theoretical core.
2.1.2 Working memory and the unattended speech effect
A collaborative research project between Dr. Pierre Salame of the CNRS Strasbourg and Baddeley has shown that immediate memory for visually presented items is impaired by the presentation of irrelevant speech (301). This effect Is not dependent on the meaning of the unattended speech, but does depend on its phonological similarity to the remembered items. The effect is abolished when subvocal rehearsal is prevented by articulatory suppression. This pattern of results has been explained in terms of the articulatory loop component of working memory. Subsequent results have shown that the effect is not sensitive to sound intensity, and does not occur when white noise is used instead of speech, suggesting the existence of some form of filtering process "that prevents non-speech material accessing the articulatory loop (302). We are exploring the characteristics of this filtering mechanism in more detail; in one study we have shown that orchestral music has a small but significant effect, whereas vocal music has a large effect. We plan to extend this research, using the phenomenon to explore the characteristics of the assumed speech filtering mechanism. We also plan to explore the generality of the impairment by looking at other cognitive tasks such as reasoning, arithmetic and comprehension.
2.1.3 Short-term memory for nonverbal sounds
Logie and Edworthy are investigating the overlap between mechanisms involved in the processing and storage of verbal and musical material, within a working memory framework (201 U). Previous experiments by Baddeley & Lewis have shown that rhyme judgments appear to be unaffected by articulatory suppression, suggesting the possible involvement of an auditory image, rather than subvocallsatlon. Experiments in this project so far suggest that pitch discrimination is disrupted by simultaneous rhyme judgments, but not by articulatory suppression. Melody memory appears to be disrupted by both of these secondary tasks. Neither pitch discrimination, nor melody memory, were affected by a control secondary task involving comparisons of visual symbols. These results suggest that melody memory involves both the articulatory loop and auditory imagery, whereas pitch discrimination involves only auditory imagery.
It appears that concepts developed within the context of verbal short-term memory can successfully be applied to the study of auditory, nonverbal information. The working memory framework will continue to be used in further experiments on this project.
2.1.4 Short-term memory for visuo-spatial material
The working memory concept has been applied successfully to an understanding of the short term storage and processing of verbal material. It has also been applied to the study of similar functions for short term visuo-spatial material, with some measure of success. However, the 'visuo-spatial sketch pad' of working memory is less clearly understood, and the experimental techniques available lack the tractability of techniques such as articulatory suppression or unattended speech that have been used in the study of the articulatory loop.
Logie has investigated the effects of unattended visual material on the use of a visual imagery mnemonic for storage and recall of word lists. This unattended material was found to disrupt visual mnemonic performance, but did not affect verbal rote rehearsal. This suggests two main conclusions: despite instructions to ignore the visually presented material, it is processed at a level sufficient to disrupt visuo-spatial cognitive function; second, unattended visual material appears to provide a more tractable technique than was previously available, for the study of visuo-spatial processing within the framework of a working memory.
2.1.5 Visual working memory
A series of experiments by Idzikowski and Baddeley has explored the hypothesis that eye movements may play a crucial role in the setting up or maintainance of visual images (8). An Immediate memory task based on Imagery was shown to be disrupted when subjects were required to track a moving point of light, while no disruption occurred when involuntary eye movements were produced as a result of post-rotational nystagmus. We are currently Investigating whether disruption is caused by the voluntary eye movements themselves, or by the need to control the movement of visual attention.
2.1.6 Working memory and reading
Research on adult fluent reading has confirmed that differences in comprehension ability can be predicted on the basis of a measure of working memory capacity (23), while other studies have attempted to examine the role of the articulatory loop system and phonological coding in reading. They suggest the possible existence of two separate systems, one dependent on subvocal articulatory coding, the other reflecting the separate use of an acoustic image. Evidence suggests that the subvocal articulatory loop system is important for certain reading tasks where a high degree of accuracy is required, and may also play an important role in learning to read (16).
Collaborative work between Logie and Baddeley and Ellis and Miles of U.C.N.W. Bangor suggests that children with specific reading difficulties typically show impaired verbal memory span. They show every sign of using the articulatory loop component of working memory, but the capacity of this system seems to be impaired. Studies of children with developmental reading disabilities in a school specialising in dyslexia so far suggests a surprising degree of uniformity of deficit. A comparison of pattern of errors in these children suggest that earlier claims that they resemble adult deep dyslexics are unfounded, with the pattern resembling more closely that of adult surface dyslexics. Children were often observed to attempt to read by sounding out and remembering each letter of a new word, before attempting to blend the sounds into the target word. What we know of working memory suggests that this is likely to be a highly inefficient strategy, and an alternative approach has "been suggested based on encoding the sound of consonant-vowel pairs or clusters (6; 12). Finally, collaborative work with the Rivermead Rehabilitation Centre Oxford, has looked at the remediation of acquired dyslexia; results suggest that although progress may be slow, relearning is certainly possible (409).
2.1.7 Working memory and retrieval from long-term memory
A series of 9 experiments has studied the effect of a demanding secondary task on learning and retrieval from episodic and semantic memory. While concurrent tasks clearly impaired learning capacity, they had little or no effect on the probability of retrieving an item from long-term memory, although latency effects were detected. These results suggest that the process of "searching" long-term memory is relatively automatic and does not require attentional capacity (21).
2.1.8 The role of memory in speech perception and production
Gathercole plans a series of studies to investigate the role of the mechanisms involved in speech perception, speech production and reading on the short-term retention of speech material. Two studies exploring the influences of judgments of phonology and speech output preparation on short-term memory are in progress, and some interesting results are emerging. Further work will in particular consider the developmental relationship between short-term memory and other speech-based cognitive skills.
Z.1.9 The neuropsychology of working memory
Patients with grossly impaired memory span but normal speech and normal long-term memory have presented an important challenge for memory theory in recent years. Baddeley has studied one such patient in collaboration with Dr. Giuseppe Vallar of the University of Milan and they have shown that a working memory interpretation can account for the data by assuming a deficit in the phonological storage component of the articulatory loop (337). Subsequent work has used the occurrence of this very pure deficit to look at the role of that component of working memory in other tasks. One series of experiments has shown that the patient is capable of making phonological judgments accurately, and of comprehending most spoken and written text. She does however encounter problems under certain conditions when the memory load of comprehension is relatively great (338). We are continuing to study the reading comprehension of this patient since it shows certain anomalies that are potentially of considerable theoretical Interest.
A further series of experiments was concerned with the question of whether the articulatory loop system depends on the capacity for overt articulation. A group of anarthric patients without the capacity to articulate were studied by Baddeley in collaboration with Barbara Wilson of the Rlvermead Rehabilitation Centre in Oxford. Their memory span performance was virtually normal with clear evidence of phonological coding and subvocal rehearsal, indicating that the articulatory loop system does not require peripheral articulatory feedback in order to operate. There is however clear evidence that memory span is grossly impaired in many aphaslc patients and Gathercole and Baddeley propose to use the techniques developed within working memory to study this.
2.1.10 Working memory and dementia
Baddeley and Logie have been invited to participate in a major project on senile dementia of the Alzheimer type by Drs Spinnler and Della Sala of the Department of Neurology at the University of Milan. Current experiments are exploring the hypothesis that patients suffering from dementia show a particular disruption of the operation of the central executive component of working memory. PhD research at the APU by Morris (236) has shown that mildly demented patients show every evidence of using the articulatory loop system, but are nevertheless impaired in overall memory performance.
Similarly, collaborative work with Milan has shown that the immediate recall of unrelated words by demented patients shows a relatively normal recency effect, coupled with grossly impaired performance on earlier items and grossly impaired verbal and spatial memory span (319 U). We are currently carrying out a series of experiments concentrating on the scheduling and time-sharing role of the central executive component of working memory. Preliminary results suggest that even mildly demented patients show excessive difficulty in scheduling the simultaneous performance of two tasks. We plan to continue research in this area with two aims in view, first to identify sensitive tasks for use in a dementia screening battery, and secondly attempting to use data from patients to extend our understanding of working memory.
2.2 STUDIES OF LANGUAGE AND SPEECH (Project 13) (Anderson, Black, Cutler, Johnson-Laird, Norris, Silverman)
2.2.1 Prosody
Cutler has investigated the role of prosodic information in recognising words, and has found that although lexical stress patterns are lexically represented (B5), lexical stress does not constrain lexical access (85; 84 U). Instead, stress rhythm appears to play a crucial role in pre-lexical segmentation rather than in word identification (254; 84 U). Higher levels of prosodic structure such as sentence accent and prosodic marking of discourse structure are perceptually valuable.
Prosodic structure carries information about a number of levels of organisation in speech. Silverman is developing a model of how the fundamental frequency (or pitch) is generated, and implementing this model in a computer speech syn the sis-by-rule program. The model views pitch contours in speech as being the result of a hierarchy of factors. At the lowest level, rules remove the "mechanical" sound of computer speech and increase the clarity by representing the influences of vowels (314) and consonants (315) on pitch contours. The middle level is based on current models of intonational structure, and generates a repertoire of intonation contours that is at the same time both richer and phonetically more accurate than existing synthesisers. Collaborative work with Ladd and Scherer in Giessen has shown that intonation contour type and range function independently in communication, and so ought to be separately represented in any model of the production of prosody. Therefore rules in the highest level constrain the overall range of these contours to make them appropriate to a discourse context.
Studies of pre-lexical segmentation by Cutler and Norris have shown that the phonological structure of one's native language can affect the strategies one uses in segmenting continuous speech; for instance, French speakers appear to segment speech into syllable-size units whereas English speakers do not (87).. However, further collaborative studies with Mehler and Segui in Paris cast doubt on the concept of language-specific levels of representation in speech processing: Norris and Cutler have accordingly proposed a new model of pre-lexical segmentation in continuous speech recognition (254), which proposes a language-universal strategy of searching for potential boundaries. The application of this strategy to languages of different phonological structure has consequences as diverse as the exploitation- of stress rhythm (in English) and apparent syllabification (in French).
Similarly, phonological structure of dialects can affect speech perception strategies within a single language: with D.R. Scott, Cutler found that speakers of British English can acquire and perceptually exploit contrasts in American English, but only if the contrast was not already present, serving a different function, in their native dialect (303).
2.2.2 The representation of words in the mental lexicon
It has been argued that because words with complex morphology or multiple meanings or many syntactic constraints are no harder to recognise than simple words, morphological, semantic and syntactic complexity may not be represented in the mental lexicon. Cutler has argued, in contrast, that lexical entries do contain representation of morphological, syntactic and semantic complexity (82; 83), but this does not make words harder to recognise; thus word recognition time is an inappropriate measure of the internal structure of lexical entries (82; 83). The role of morphological structure in word recognition allows an explanation of the general preference for suffixes over prefixes in the languages of the world (86).
2.2.3 Comprehension of discourse
Johnson-Laird, Black and Anderson have been engaged since the autumn of 1982 in a project on the factors that affect the comprehension of discourse. These studies are motivated by the theory that comprehension enables readers to construct a mental model of the situation described in the discourse (177). The theory predicts that two main variables should affect the ease of understanding discourse: referential coherence and plausibility. Both these variables have now been shown experimentally to have the predicted effects. A text is referentially coherent if each sentence in it makes reference, either explicitly or implicitly, to at least one entity referred to in another sentence, and hence it is possible to integrate all the sentences into a single mental model. When referential coherence is disrupted, a passage is harder to understand, takes longer to read, and is remembered less well. Conversely, even if the sentences in a text are jumbled up in a random order, referential coherence suffices to make the text considerably easier to understand. Such a text, however, obviously recounts an implausible sequence of events, and plausibility has been shown in further experiments to have a marked Influence on the ease of understanding discourse and of detecting inconsistencies in it. Experiments are in progress to determine the extent to which the two variables of coherence and plausibility interact.
2.3 STUDIES OF NORMAL READING AND WRITING (Project 33) (Cotter, Evett, Hull, Morton, Nimmo-Smith, Norris, K.Patterson, Wright)
2.3.1 Models of word naming
Norris has developed a detailed model of the word naming process and related it to other measures of word recognition (248; 255 U; 253), while Norris and Brown (253) have shown that although analogy theories and race-models are formally distinct, in practice they are likely to prove empirically indistinguishable. Norris and Nimmo-Smith (255 U) have developed mathematical models of both naming and lexical decision and have shown that the race model can account for many of the differences between the two tasks. Nimmo-Smith and Morton have developed mathematical models of response times for competing processes, which have" a much wider applicability than solely to word naming.
Two studies of normal reading by Karalyn Patterson address issues which are being concurrently tackled with data from neurological patients.
(a) Modelling the translation of orthography to phonology. The mechanisms by which a reader can assemble a phonological code (pronunciation) for a new orthographic string (e.g. a nonsense word like dake or an unfamiliar name like Tharp) continue to be a matter of compelling theoretical interest and controversy. Original assumptions of a set of a 'nonlexical' rules or mappings between letters and phonemes have been attacked by 'lexical' theories which propose that all such knowledge of mappings can be generated from segmented lexical entries. While there have Indeed been findings in the last five years which the nonlexical theory in its original form cannot explain, there are reasons (some of them neuropsychological) for preferring to retain the notion of two separable routines (one lexical or based on specific words, one nonlexical or based on mapping rules at the sub-word level) for translation between orthography and phonology. Accordingly Patterson, Morton and Nimmo-Smith have developed a modified version of the dual routine model which can deal with the recent controversial findings. A qualitative description of the model is in press (272) and a more detailed and quantitative characterisation of the model is in preparation (243 U). Whilst some of the modelling process has been able to proceed on the basis of available data, we have also been forced to collect substantial data of our own [e.g. Evett et al, (100 U)] because the complexity of our developing model requires a level of specificity not available in most previous research 1n this area.
(b) Right hemisphere reading? As mentioned in the section on acquired dyslexia, some of Patterson's recent work has Involved an evaluation of this controversial topic (232; 266; 267). One aspect of its importance concerns the putative implications of deep dyslexia for models of normal reading. If, as suggested by some theorists, deep dyslexic reading relies on right-hemisphere processes, then we can legitimately apply findings from deep dyslexia to models of normal reading only if we can demonstrate that such right-hemisphere processes also contribute to normal reading skill. We are beginning a collaboration with a team of neuropsychologists in Zurich (Drs. Landis and Regard) who claim to have evidence for such right Hemisphere participation in normal word recognition and reading.
(c) Dual-task studies of word recognition and production. Do the lexical representations underlying word recognition and production within a modality (either speech or print) constitute one sub-system or two? That is, when we recognise-a spoken word, for example, is the hypothetical unit responsible for this event the same as that which must be activated when we want to speak that same word, or are these 'speech' units to be found in a separate subsystem? This issue, although in principle addressable with a neuropsychological approach, has proved somewhat resistant to resolution on the basis of available patient data. Shallice and McLeod have developed a dual-task methodology with normal subjects to address questions of this general type, and have applied - it to the analogous issue on the auditory/speech side (311 U); a collaboration between McLeod and K. Patterson is beginning to ask parallel questions about recognising printed words and output of words in an orthographic code.
2.3.2 Reading and comprehension
A series of studies by Wright and Hull have explored the way readers process conditional instructions (Do A if B), particularly when these instructions involve negatives (e.g. Unless, If not). It has been shown that people have alternative strategies for representing the information in such conditionals (168), but that the mapping between strategy selection and conditional term tends to be stable across rhetorical context (imperative/declarative) (450 U). Specifically, readers represent sentences containing Unless as a negative action and a positive condition, whereas If not is represented as a positive action and negative condition.
Many of the dissimilarities between Unless and If not are mirrored by those between Until and While (449 U). This suggests that the negative prefix un elicits a characteristic processing strategy from readers, a strategy different from that used for not.
Drawing inferences is a normal part of most reading activities. Cotter has explored some of the determinants of inferences relating verbs (e.g. cut) to likely instruments (e.g. scissors); we have also examined where these inference processes occur within the ongoing activity of text interpretation (79 U). By noting whether or not reading was slowed by the introduction of an unexpected Item, we were able to show that for some verbs, and Indeed some nouns, readers will instantiate an inferred category at the point of first reading the word (e.g. most readers instantiate the noun nurse as female) (80). The structure of the semantic representation of the word appears to be a strong predictor of the Inferences that readers will draw.
2.4 STUDIES OF INFERENTIAL PROCESSING (Project 46) (Johnson-Laird)
Two main sorts of inference are under investigation by Johnson-Laird and his colleagues: formal syllogistic reasoning and informal everyday inferences. Experiments on syllogisms have borne out the prediction that the main factor that affects the difficulty of a deduction is the number of different mental models that have to be constructed in order to imagine all the main ways in which the premises could be true (175). This theory has been fully implemented in a computer model; the next step is to generalize the theory to all inferences involving quantifiers, and to the variety of quantifiers that occur in ordinary language. Everyday inferences are made much more rapidly and much more tentatively than formal inferences, e.g. when you learn that a suspect in a murder trial was somewhere else at the time of the murder, you are likely to infer that he is innocent. When subjects are challenged on such conclusions, they are able to construct alternative models that violate their initial conclusions. A series of experiments (in collaboration with Dr. B. Bara of the University of Milan) are underway to investigate the ease of formulating such alternative conclusions as a function of the strength of the constraints that the premises exert on the initial conclusion. A long-term, aim of this collaboration is to develop a computer model of the process of everyday inference and the role of general knowledge within it.
2.5 COGNITIVE MODELLING AND COMPUTER SIMULATION (Project 34) (Barnard, Hinton, Johnson-Laird, K.Lewis, McLeod, R.Patterson, Shallice, Young)
2.5.1 Production Systems as a vehicle for cognitive modelling
Following his earlier work on the modelling of errors in subtraction, Young has been collaborating with a number of people at the Open University on a paper based on 10 case studies of production system modelling in order to explore the limits of applicability of the approach (262 U). Two factors appear to determine the feasibility of the modelling enterprise: the quality of the subjects' performance, and the possibility of performing an adequate task analysis. The account of errors in purely mechanical terms is also being extended by some speculations as to what underlies a child's understanding (and misunderstandings) of simple arithmetical tasks (462 U). An account is offered in terms of the partial assimilation of arithmetical operations by multiple schemata, each of which can contribute to an aspect of the solution. In favourable cases this can lead to the child's "invention" of the solution to a novel kind of problem.
As well as these particular models for particular tasks, there is interest in the notion that the Production System "architecture" itself is responsible for certain features of human cognition. A speculative paper (456) makes the case that properties of the architecture give rise to such invariants of cognition, as that problem solving occurs as heuristic search in a problem space, or that long-term memory search takes place in cycles Of retrieval and re-descript1on.
2.5.2 The functional architecture of cognition
Shallice has been engaged in two types of study concerned with this issue. The first in collaboration with McLeod and Lewis is to attempt to develop dual task procedures to answer theoretical questions about functional architecture that have arisen from neuropsychological work. The issue on which there is neuropsychological controversy that they tackled first concerns whether the structures responsible for morpheme-level phonological processing (i.e. logogen or word-form systems), are common to speech input and output. They found that using two tasks that stressed word perception and word production (i.e. detecting a name In a string of auditorily presented words and reading aloud visually presented words both presented at a rapid rate, subjects were able to combine the two tasks making only 10% more errors on average than on the two tasks presented in isolation. They buttressed their conclusion that separate phonological processing structures are involved in speech perception and production by a series of control experiments designed to facilitate theoretical extrapolation from the results of dual task experiments to the nature of the underlying cognitive structure. A paper is being revised for publication (311 U).
In a theoretical paper Shallice (307) assessed Fodor's (1983) theory that the brain is composed of modular input systems and equipotential central systems. He was able to show that modularity as defined by Fodor was too restrictive a concept to give, a plausible characterisation of the architecture of all perceptual systems but also that Fodor's arguments that central cognitive systems are equipotential are probably not valid.
Theoretical approaches within APU often seek to furnish accounts of cognitive phenomena on the basis of partitioning the human cognitive system into a variety of processing and memory subsystems. Work on one such model Barnard's "Interacting Cognitive Subsystems", has been developed to provide a conceptual framework for analysing a variety of cognitive tasks. The initial emphasis was to provide accounts of a broad range of short term memory phenomena on the basis of subsystems required to understand and produce language (32). The general purpose architecture, developed primarily to account for laboratory phenomena, is now being explored in relation to its possible applications to the more complex and practical problems of understanding and memory in the context of human-computer dialogues (33 U).
2.5.3 Parallel models of memory and perception
Current computer architecture is essentially based on serial processing. It seems likely that the brain approximates more closely to a parallel processing system. In some of the most original theoretical work carried out at the APU during this period, Hinton developed and simulated a parallel processing model and applied it to memory, perception and motor skills (144; 145; 146; 147; 148). The work was started in the rather more computer-rich departments of Sussex and San Diego, and is now bearing fruit in the computer science department of Carnegie-Mellon University, but was carried on successfully at the APU by using, with great ingenuity, the very spartan technical resources of an Apple Microcomputer. This line of research holds great promise, both in implications for understanding human cognition, and in its potential significance for computer design.
2.5.4 Simulation of musical Improvisation
Johnson-Laird has written a suite of programs that simulate the process of musical improvisation. The basic theoretical claim on which the programs rest is that the process of improvisation, since it occurs rapidly and fluently, must minimize the amount of musical structure that has to be constructed in order to produce each note. Hence, the required computational power may be as weak as that of a finite state device. Johnson-Laird has devised (a) a program that improvises a bass line to a given chord sequence, and (b) a program that Improvises a melody to a given chord sequence. Both programs utilize merely this degree of power yet achieve considerable verisimilitude. In the case of the generation of chord sequences, however, there is no such real-time constraint, and in fact it appears that the power of at least a push-down automaton is required to carry out this task. A third program has been devised that creates acceptable tonal chord sequences, and it depends on this greater degree of computational power. These musical programs have been interfaced with the Unit's specially built computer for generating complex sounds; and Roy Patterson and his colleagues have developed the software for producing musical notes. This collaboration will continue, and the project aims to lead both to theoretical insights and to practical applications in the domain of musical technology.
2.6 NEUROPSYCHOLOGY OF COGNITIVE BREAKDOWN (Project 1) (Baddeley, Duncan, Gathercole, Harris, Hatfield, Kay, McCarthy, Marcel, Morton, K.Patterson, Shallice, Sunderland, Wilkins)
2.6.1 Cortical blindness
Marcel and Wilkins are continuing work on cortical blindness. Patients with lesions to cortical visual projections on one side, but not to subcortical projections, have no phenomenal vision in one half of the visual field. However some of these patients show accurate pointing to and accurate adjustments of wrist and fingers to shape, size and orientation preparatory to grasping objects in the blind field of which they are unaware. In addition some patients are sensitive to the structural descriptions of strokes which constitute letters. Further, in the blind field conscious vision of after-images and of illusory contours can be produced if facilitated by related stimulation in the sighted field. Such findings not only elucidate the nature of cortical blindness, but throw light on the nature of visual consciousness (228 U).
2.6.2 Action and Motor Control
Although almost completely ignored by the research literature, it is a common-rehabilitative technique (for eliciting and improving motor control for both peripheral and central problems) to embed the required behaviour in a higher-level task. In conjunction with the Rehabilitation Centre at Addenbrooke's Hospital Marcel has been exploring the validity of the phenomenon and in whom it occurs. We have established in a sample of neurological patients with various motor deficits that if a piece of motor behaviour is part of an action, then it is easier to retrieve and is performed more efficaciously than when it is the nominal focus of the task. This appears not to be entirely an attentlonal phenomenon. The importance of this is twofold. Therapeutically, one may validate and delineate a technique for eliciting and improving motor control. Theoretically, one may throw light on the way in which motor behaviour is influenced by the nature of the intention.
2.6.3 Acquired dyslexia
With realisation of the multiple sub-processes involved in a complex cognitive skill like reading, it becomes obvious that neurological damage will yield a wide variety of patterns of breakdown rather than just a few discrete syndromes. This state of affairs underlines the value of the single case-study approach, so extensively adopted in research at the APU and gaining increasing acceptance in neuropsychological centres elsewhere. Despite the fact that each single case is unique, basic patterns ('syndromes' or 'symptom complexes') provide a framework within which the variability can be described. Review chapters summarising the most prominent patterns of acquired dyslexia (268) and the major neuro-psychological tests and techniques for assessing acquired reading disorders (74) will appear shortly in a major handbook on aphasia.
2.6.4 Non-phonological reading
Previously, much of the Unit's research in this area concentrated on deep and phonological dyslexia, involving disorders of the phonological sub-processes which enable normal readers to pronounce unfamiliar printed words or names. The only recent work on deep dyslexia has been K. Patterson's critical evaluation of the hypothesis that deep dyslexia reflects reading skills localised in the right hemisphere of right-handed (left-hemisphere dominant) individuals (232; 266; 267). Phonological dyslexia is currently being studied by Marcel and Patterson in an attempt to specify why these patients, whose reading of familiar words is nearly normal, fail to pronounce even very simple new words or nonwords (like dake). At least three procedures are necessary to read aloud such a letter string: segmentation of Its orthography, assignment of phonology to the individual segments, and assembly of the phonological segments into a spoken syllable. Alternative theories about the nature of these procedures in normal reading make differing predictions regarding the locus of the deficit in phonological dyslexic patients, and also about the way in which patients with other patterns of reading impairment should perform on tests of these component procedures (182). New more sophisticated tests, devised by Marcel and Patterson, have now been given to five phonological dyslexic patients and are currently being given to other patients.
2.6.5 Phonological reading
Recently a primary focus of our work on applied dyslexia has been patients whose reading involves an abnormal reliance on the phonological processes which are absent or impaired in deep/phonological dyslexia. Although at an early stage this pattern was considered a single syndrome (called surface dyslexia), it is now clear that phonological reading can take a variety of forms. Indeed, Shallice and McCarthy (310) argue that the form of reading impairment shown by the original patients labelled surface dyslexic is best conceived not as reliance on a phonological routine but rather on a compensatory strategy related to letter-by-letter reading.
Two issues of major theoretical interest have guided out work on phonological reading. The first concerns its informative value for theories of the phonological routine in normal reading, requires a relatively pure disorder as in the patient studied by Shallice, Warrington and McCarthy (313). This patient was grossly impaired in terms of reading based on recognition of whole familiar written words but had a largely preserved phonological routine for reading as demonstrated by (a) normal reading speed; (b) a large effect on probability of oral reading success of the word's regularity of grapheme-phoneme correspondences; (c) a predominance of errors of a pure regularisatlon type (e.g. pint: "pint", rhyming with "mint"); and (d) equivalent success on nonsense words and regular monosyllabic words. The 'multiple levels' model of phonological reading suggested by Shallice et al (313) has been further developed by Shallice and McCarthy (310). The model assumes that information about different sizes of orthographic unit - grapheme, demisyllable, syllable and morpheme - can be passed in parallel from orthographic to phonological processing systems in spelling-to-sound translation. Impaired phonological route reading corresponds to a loss of some - normally the higher - of these levels of correspondence. Analogies with scene recognition programs were developed to support the idea that the model was computationally practicable.
The second issue concerns the status of reading comprehension in patients whose oral reading relies on phonological processes. Original interpretations of surface dyslexia concluded that both pronunciation and comprehension were dominated by segmental phonological assignment (so that, for example, the printed word bear would not only be pronounced as "beer" but also understood as referring to an alcoholic drink rather than a large mammal). Now, however, and entirely consistent with our process models of reading, patients have been identified who read (aloud) phonologically but comprehend orthographically (saying "beer" to bear but defining the written word correctly) (230; 183). Phonological reading associated with deficits at output rather than input does predictably co-occur with output problems in other language tasks as well, particularly spontaneous speech and naming. Marcel's study (230) also focussed on the variability of the patient's responses and concluded that observed behaviour patterns, rather than reflecting an information-processing deficit, may reflect strategies or biases in response to a deficit. These biases appear to be affected by attention and task pragmatics.
A book on the topic of phonological reading, edited by K. Patterson, Marshall and Coltheart (271) will be published in the spring of 1985 and includes four chapters written by APU members (230; 272; 183; 310).
2.6.6 Letter-by-letter reading
In earlier papers, Warrington and Shall1ce (1980) and Patterson and Kay (270) have presented evidence that patients suffering from the classical syndrome known as letter-by-letter reading show no comprehension or ability to perform lexical decision on briefly presented words. Reading is dependent on a laborious letter-by-letter process. In collaboration with Saffran, Shallice has undertaken a lengthy series of experiments on an apparently very typical letter-by-letter reader who performs well above chance on lexical decision tasks and semantic judgements even on words he can neither read nor identify. He also appears insensitive to the appropriateness of affixes when performing lexical decision. A number of possible explanations of why he differs from letter-by-letter readers are developed. A short account of the work 1s presented (309), and a core extensive publication is at present being revised (312 U).
2.6.7 Reading, naming and semantics
It is a moot issue to what extent the semantics accessed by written words and by perceived objects are common or separate. Marcel, in collaboration with Margolin of the Good Samaritan Hospital, Portland, Oregon, has studied a woman who suffered anomia and alexia following a head trauma. As her language improved, she produced descriptive circumlocutions when both reading and naming. These responses were strikingly similar for a given word, whether in reading or naming a picture. This observation together with experimental interactions between semantic, phonological pictorial and orthographic information suggest that a common set of semantic representations are accessed and used in recognition and naming of objects and in recognition and reading of words. (229 U).
2.6.8 Acquired dysgraphia
Disorders of writing and spelling are rapidly catching up with reading as a major focus of neuropsychological research. It is particularly Intriguing to identify those aspects of spelling procedures which can be modelled with processes or routines borrowed from models of reading as opposed to those aspects of reading and spelling which require differing theoretical accounts (140; 264). In terms of patterns of acquired dysgraphia, as with reading, Patterson and Hatfield now have considerable evidence about spelling performance both where the patient has severely impaired or abolished phonological skills (141; 274) and where, due to impairments in the word-specific lexical procedures for spelling, the patient depends primarily upon phonological skills (140).
2.6.9 Aphasia
The period covered by this report saw the termination of Patterson and Morton's DHSS-supported research to evaluate techniques of speech therapy for word-finding difficulty in picture naming tasks. Summarised briefly, the study demonstrated that (a) single applications of techniques based on the sound of the sought-for word provide major immediate assistance to the patient but no lasting benefit (Patterson, Purell and Morton, 1983); (b) single applications of techniques based on the meaning or referential properties of the word provide facilitation of word-finding that may last hours or even days (154); intensive treatment sessions (every day for one or two weeks) provide significant cumulative benefit whether the techniques used are based on word sound or meaning; but naming performance begins to decline., as soon as the therapy period terminates (155 U). The very extensive data collected also enabled us to address other questions of interest, such as consistency of naming performance over sessions (153). Some of the therapeutic implications of this study are addressed in these published papers, and also in our final report to the DHSS (273). Also, we are exploring implications for design of therapy procedures derived from the single-case methodology which has been so theoretically productive (269 U).
Work is beginning on the aphaslc disorder (or set of disorders) known as agrammatism, in which language production and/or comprehension can be roughly characterised as impaired at the level of grammar or syntax. One case study of a pattern rarely reported in the literature has revealed a patient agrammatlc in spontaneous speech but not in other tasks requiring verbal output such as repetition or oral reading of sentences (265 U). Other work just beginning (K. Patterson in collaboration with L. Tyler) will use Dr. Tyler's sophisticated model of word recognition in speech comprehension to characterise various aspects of agrammatlc disorders.
2.6.10 working memory and aphasia
A new line of enquiry by Gathercole and Baddeley concerns the memory characteristics of neuropsychological patients with aphasia. Research using normal populations of subjects has established a strong articulatory component In short-term memory, as demonstrated in particular by word length and articulatory suppression effects. Recent work by Baddeley has shown that dysarthric patients - people who have motor problems associated with articulation - still display evidence of articulatory coding in memory. The present study seeks to investigate the contribution of speech coding to the short-term memory of some aphasic patients who have speech output problems in the absence of dysarthria. It is hoped that this investigation will provide further insight into the nature of articulatory representations in memory. Close collaborative contacts with the Speech Therapy Department at Addenbrooke's Hospital in Cambridge have been established, and memory testing of aphasic patients is in progress.
A further project planned by Gathercole and Baddeley concerns memory in children with language disorders. The children in whom we are particularly interested spontaneously produce utterances of limited length only, and one hypothesis which will be investigated is that an Impaired auditory memory system is associated with this type of disorder. More generally, the aim of this investigation is to explore the reciprocal relationship between short-term memory and the development of language. At present, we are involved in establishing a suitable group of children for testing.
2.6.11 Memory span and epilepsy
Wilkins, in collaboration with colleagues at the Instituut voor Epilepsiebestrijding, Heemstede, The Netherlands, has adapted verbal and non-verbal tests of short-term memory for clinical use with epileptic patients. The tests take the form of television games that are entertaining when played for lengthy periods. They have examined patients whose seizures are reasonably well controlled but whose medication is insufficient to eliminate occasional epileptiform EEG abnormalities. The tests have revealed selective. Impairments of verbal or non-verbal memory span associated respectively with focal EEG abnormalities of the left and right temporal regions (1; 2).
2.6.12 Amnesia
Recent research in this area by Baddeley concerned the question of whether amnesic patients show impaired episodic memory but normal semantic memory (10), a question that led on to the exploration of autobiographical memory in amnesia. Collaborative work with Barbara Wilson of the Rivermead Rehabilitation Centre Oxford has led to the observation that marked differences exist in the extent to which otherwise apparently equivalent patients have access to memory of. their earlier lives. This led to an initial triple categorisation of patients as having normal, clouded or confabulatory autobiographical memory (27). Patients who confabulated all suffered from frontal amnesia; a further study .comparing frontal amnesia with other amnesic deficits 1s in.preparation. We are also in the process of developing an improved assessment of autobiographical memory (Jointly with Wilson, Rivermead, Oxford and Kopelman, Institute of Psychiatry, London).
2.6.13 Studies of semantic memory
The first detailed quantitative case studies of a selective impairment of certain semantic classes of material has been undertaken by Warrington and Shallice (339). Their paper concerned the visual identification and auditory comprehension deficits of four, patients who had made a partial recovery from Herpes Simplex encephalitis. In all four patients a very large difference between the ability to identify Inanimate objects and the Inability to identify living things and foods was demonstrated. In two patients it was possible to compare processing of auditory verbal and visual non-verbal stimuli; a similar pattern of dissociations was seen in both modalities. In one patient comprehension of abstract words was far superior to that of concrete words. The findings are interpreted in terms of the category specific organisation of semantic systems.
In a theoretical review Shallice (309) has analysed disorders within the semantic system on two dimensions in addition to the possible category specificity of the impairment - according to the modality of presentation (or of connections within the semantic system) and on whether the disorders are of "access" or "degradation", developing a dichotomy originally put forward by Warrington and Shall1ce (1979).
2.6.14 Amnesia and everyday memory
Our previous progress report referred to an ongoing study by Sunderland, Harris and Baddeley concerned with the measurement of everyday memory problems encountered by patients suffering from closed head injury. That study was completed, (320; 321) and an equivalent study on the memory problems of the elderly has subsequently been run. The main focus of both studies was on the relationship between laboratory-based objective memory tests and memory problems encountered in everyday life. We explored the use of Interviews both with the patient and with relatives, together with the use of check lists and diaries. The results of both studies indicated that neither standard tests nor questionnaires are entirely satisfactory. Some standard tests reveal clear deficits that appear to be unrelated to everyday problems. On the other hand, interviews with the patient proved to be of limited value; the patients often appear to forget their lapses of memory (321; 24). This has led to a collaborative project with Barbara Wilson of the Rivermead Rehabilitation Centre concerned with evaluating a behavioural memory test. This comprises a number of objectively scoreable subtests, each of which attempts to simulate one area of everyday life that patients report gives rise to memory problems. Performance on the test 1s currently being validated against extensive observations by therapists, and will be correlated with objective memory tests, and subjective estimates by the patient and a relative. Preliminary results are encouraging (410), the test is already arousing considerable interest and being used in joint studies on dementia (with Dr. Wegener of the Merck Drug Company in Darmstadt), on depression and ageing (with Dr. Poon In Boston) and on memory in stroke patients (with Dr. Lincoln in Nottingham).
The head injury and memory project stimulated an interest in treatment. A preliminary survey indicated considerable interest but relatively little activity in treatment of memory problems (136). This in turn led to the organisation of a workshop, jointly with a number of clinicians interested in this topic. The publication of a book on the management of memory problems followed, with chapters contributed by the APU on the relation between theory and therapy (11), and on methods of improving memory (134). A second conference was organised some two years later, and it was gratifying to note the extent to which this complex but important area is developing. While it would be irresponsible to pretend that dramatic improvements in memory problems can be achieved, there is good evidence that given appropriate techniques some amelioration is possible.
2.6.15 Frontal Lobe Disorders
Five empirical studies of frontal lobe function have been carried out:-
(a) A study of planning abilities was carried out by Shallice and McCarthy using a test devised so as to be related to the Tower-of-Hanoi puzzle, but to be of graded difficulty. In an unselected series of localised lesion patients it was found that the group of patients with left anterior lesions were significantly impaired when compared with either a left posterior lesion group or. a right anterior lesion group. In an experiment on normal subjects it was found that articulatory suppression had no effect on the ability to solve the task and an "Inner speech" deficit did not therefore seem a plausible explanation of the left anterior deficit. A short account of the research has been published (306); a more extensive version is being prepared for publication.
(b) A directed forgetting task was developed by Shallice and McCarthy in which a form of span involving subsequent Instruction to forget some of the presented material Is contrasted with normal span. Patients with anterior lesions were selectively impaired on the directed forgetting condition. As no interaction was observed with hemisphere the deficit would appear to arise from a general programming or attentlonal problem for a novel and demanding task rather than any difficulty with, say, inner speech. It is at present being considered whether, further investigations need to be carried out on this point before the work is written up for publication.
(c) Previous work at the National Hospital with Professor Warrington and Ms Oldfield has attempted to develop a battery of tests sensitive to frontal lobe lesions. For a number of methodological reasons this work has not been successful. Shallice and Lewis together with Warrington have been engaged in an attempt to improve their previous technique of using an unselected series of patients with localised lesions by screening such patients extensively for inclusion in a further series. Only patients in whom basic perceptual and cognitive skills were intact and IQ measures are relatively unimpaired are being included; the patients are then being assessed on a group of ten tests thought from the literature to be sensitive to frontal lobe pathology. The series is at present about half complete and is expected to take another year.
(d) Shallice and McCarthy have extended a study originally carried out in Montreal by Wilkins and have replicated the original findings with a very different group of neurological patients at the National Hospital, Queens Square, London. In both studies, patients with frontal lesions were less able to perform a simple routine counting task, but only when it was given at slow presentation rates, a condition that required patients actively to concentrate their attention (389 U).
(e) Shallice, together with Beauvols and Oerouesne of INSERM 84, Hopital Salpetriere has been engaged in the analysis of the memory disorder of a patient showing amnesia with confabulation following operation for an anterior communicating artery aneurysm. The investigation involved an extensive series of tests of memory, attention and cognition and the provisional findings are that a frontal memory disorder exists which is not secondary to a primary amnesia, concentration difficulties or the more general cognitive problems associated with frontal lobe pathology. A preliminary account has been submitted to the International Neuropsychology Society European meeting for 1985 but investigations are still continuing.
Shallice, together with D.A. Norman of the University of California has developed a theory of the higher level cognitive function associated with the frontal lobes which is the first to be put forward within an information-processing framework. They have been able to interpret a number of impairments obtained with frontal lobe lesions within the theory (306). It is also possible to account for a number of phenomena in the psychology of attention within their approach (245). Baddeley has utilized the Shallice and Norman model within the central executive component of his working memory framework (8 U).
Duncan is investigating the relationship between the "frontal lobe syndrome" and the aspect of individual differences in the normal population measured by so-called "Intelligence" tests. People with low scores on such tests seem to have particular difficulties when beginning new tasks, yet to perform reasonably once the correct structure of operations has been achieved. In particular, they can show initially a pattern of behaviour quite reminiscent of one aspect of the frontal lobe syndrome: A failure to correlate performance with goals and/or instructions, though these have been understood verbally. It is proposed that this type of correlation Is vital when new tasks are begun, but then rapidly loses importance, perhaps even after only one or two successes. Work is in progress to test this Idea both in the normal population and In frontal lobe patients.
2.7 COGNITIVE ASPECTS OF DYSFUNCTION AND TREATMENT IN PATIENTS WITH PSYCHIATRIC AND OTHER EMOTIONAL PROBLEMS (Project 44) (Broadbent, Levey, A.McLeod, Marcel, Nulty, Sharrock, Trezlse, Watts, Williams)
This project, which has developed since 1981, has applied techniques and theories derived from experimental cognitive psychology to emotional disorders. Though clinicians have become increasingly concerned with cognitive aspects of emotional disorders, the experimental approach taken in this project is relatively innovative. The project has three general alms (a) to describe and develop a theoretical account of cognitive processing in emotional disorders (b) to evaluate cognitive change in established behavioural therapies and, where appropriate, to develop therapeutic applications of processing strategies and (c) to contribute to the general development of cognitive theory an awareness of how cognition and affect interact. The main focus so far has been on the first of these; the other two will come to greater prominence as the project develops. The subjects studied so far have been partly clinical groups (depressed psychiatric patients and overdose patients) and partly volunteers recruited by advertisement (spider phobics and 'worriers').
2.7.1 Cognitive processing in spider phobia
Spider phobics have been a major focus of work in the early stages of the project, because the discrete nature of the emotional stimulus makes the study of cognitive processing relatively straightforward. A preliminary step was to develop a questionnaire measure of spiders which both provides a rapid check on the overall severity of the phobia and also measures specific dimensions derived from factor analysis (vigilance, preoccupation and avoidance) likely to be relevant to cognitive processing (371). A Stroop test has been developed requiring the colour naming of spider words, for which spider phobics show very long response latencies (375 U). This has shown promise as an outcome measure in evaluating the effects of desensitization. Because the size of the effect is larger than in any other emotional Stroop so far reported it lends itself to studies of the processes involved, and work on this is in hand.
Another series of studies by Watts has used dead spiders as the stimuli and demonstrated a tendency for phobics to show 'shallow' processing of spiders, reflected in poor memory for them. It is planned to study the extent to which this Interferes with the clinical effectiveness of exposure-based treatment of phobics, and this will probably be studied in agoraphobics. Memory for spider words has been studied by Watts and Trezise and it has been found that spider phobics show reduced free recall for spider words, though recognition memory remains good (376 U). This is contrary to the well established finding of enhanced recall of negative words in depressed mood states, and raises issues about which disorders show enhanced and which reduced recall of salient words. A parallel study of sub-clinical anorexics is in progress, together with further work on spider phobics aimed at disentangling possible explanations of the effect. Other work has used self-report measures of cognitive representations and shown that, in phobics, spider constructs are unusually highly correlated (372) and spider imagery is poorly elaborated (374 U).
2.7.2 Concentration and memory in depression
Another line of work by Watts, Trezise and McLeod has Investigated concentration and memory problems In relatively severely depressed patients. An initial study (373) used a structured interview to explore the phenomenology of concentration problems. It also used a variety of objective tests to establish a dimension of concentration problems that spanned data domains, and yielded a representative index that could be used in future work. Current work is examining the hypothesis, derived from the preliminary study, that depressed patients have two kinds of concentration problem, one relating to processing input, and the other experienced in interactive tasks. Processing strategies designed to help patients with the first of these are currently being evaluated. Work is also in progress on the locus of memory problems in depression.
2.7.3 The assessment of specific psychological factors affecting recovery from self-poisoning
Parasuicide is a large and increasing problem, the number of incidents having increased over the past twenty years from 30,000 to 130,000 pa in the UK. Yet to date mental health workers have not found any physical or psychological treatments which reliably reduce the risk of repetition (402). This project has been set up to examine individual differences in the process of recovery from the emotional stress surrounding the attempt. Specifically, Williams and Broadbent are studying the interrelationship between mood, hopelessness and biases in attention and mood related autobiographical memory. Two independent forms of mood-related biases in processing have been identified. The first related to increased sensitivity to constructs related to negative themes (407 U) the second related to retarded retrievability of positive personal memories (406 U).
2.7.4 The effect of depressed mood on lexical decision and subsequent recognition ot emotional and neutral material
In this project, Williams examines' predictions from associative network theory of emotion by studying the effect of level of depression on the extent to which negative words are responded to more quickly in a lexical decision paradigm. Two further aspects are being studied: (a) the effects of priming on lexical decision and (b) the extent to which target words from the lexical decision task are accurately recognised later (using a false recognition paradigm).
2.7.5 Worrying and working memory
Williams, Watts and Levey are studying the effect of articulatory suppression on latency to recruit or dismiss "worrying" and "neutral" thoughts in a clinical subpanel of "worriers". This study is also indicating a distinction between several dimensions of worrying; we plan to examine the correlations between these dimensions and depression, general trait anxiety, frequency of automatic negative thoughts, and dysfunctional cognitive style.
2.7.6 Conscious and nonconscious processes in dynamic and cognitive psychology
In cognitive psychology there has recently been a resurgence of interest in both consciousness and nonconscious processes. Dynamically oriented theorists and researchers have looked to this work for its potential relevance to psychoanalytic issues. Marcel has examined this attempted rapprochement and has argued that the work in cognitive psychology is irrelevant to psychodynamic concerns. However It is pointed out that the hermeneutic focus of much psychoanalytic work is an important aspect of cognition omitted by the natural science approach of cognitive psychology.
The treatment of meaning by cognitive psychology and its significance for cognition can be helped by the work of social anthropologists (226).
2.7.7 Clinical psychophysiology
The study of cognitive functions tends by its nature to pay less attention to other functional activities Including motivational and emotional components of behaviour. This imbalance is redressed in the program of the Unit, through an area of research broadly classified as psychophysiological, which studies autonomic functions in relation to cognitive activity. This area includes studies by Levey in collaboration with Dr. Martin of the Institute of Psychiatry of classical conditioning of simple reflex behaviours which are affected by individual differences in arousal (186; 187; 188) or of more complex behaviours directly involving emotional attitudes (190). The classical conditioning paradigm can be regarded as a very simple form of Information processing in which the significance of novel stimuli 1s modified in the context of phylogenetically primitive emotional reactions. For this reason it offers an ideal focus for studying the interaction of cognitive and emotional components in laboratory based behaviours (192 U) but also offers the prospect of a better understanding of clinically relevant (maladaptive) behaviours in real life (189; 191).
A major focus of this project is the application of the concepts and methods of classical conditioning and the study of psychophysiological disorders. Studies of heart rate, pulse volume, respiration, gastric activity, and electrodermal responding have shown that some individuals fail to extinguish responses conditioned normally in these organ systems (193). These individuals are characterised by low arousal levels and by inefficient cognitive processing of environmental stimuli (194). They are consequently at risk for the development of psychosomatic disorders. The practical implication of these findings is that the methods of deconditioning and cognitive restructuring which has been successfully applied to other learned maladaptive behaviours may also be applicable in modified form with psychosomatic illness.
Other sections in the 1981-1984 report
2. COGNITIVE PSYCHOLOGY
3. COGNITIVE ERGONOMICS/APPLIED COGNITIVE PSYCHOLOGY
4. HEARING
5. MOTOR SKILL AND ACTION
6. VISUAL PERCEPTION
7. PSYCH0PHYSIOL0GY SECTION