You are in: Home » History of the Unit
PERCEPTION AND ACTION
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.
Allerhand 4.5, Carlyon 0.5, Goodrich 0.75, McKeown 2.0, Maylor 0.73, Nimmo-Smith 0.75, R Patterson 5.0, Sellen 0.73, Thurston 2.0, Tresilian 2.75, Wilkins 5.0, Wing 5.0, Allison (HSO) 4.0, Chronicle (SO) 0.5, Datta (SO) 1.08, Jeanes (SO) 1.25, McLachlan (SO) 0.25, Milroy (HSO) 1.0
Total Person Years: Scientists 29.7; Research Support 6.3
Abstract
Objectives
This programme employs psychophysical, computational and electrophysiological methods to increase our understanding of human perception and action. The objective is to develop theoretical models for perception and action in normal and impaired performance, and to apply these models to medical and communication problems.
Scientific progress and achievements over the past five years
Progress in the last five years has occurred on three broad fronts: audition, vision and movement control. In audition, a computational model (AIM) has been developed to simulate the auditory images we hear when presented with everyday sounds and to provide a basis for source identification. Applied work has led to the provision of auditory warnings in a variety of high-workload environments. In vision, an electroencephalographic investigation of photosensitive epilepsy has led to a general, unified theory of visual stress. The theory encompasses photophobia in migraine and perceptual distortions in dyslexia. It has been applied to recent developments in lighting and visual displays. In movement control, models of the variability of elemental aspects of movement (force and time) and of visual control of action have been proposed and tested. Various tasks, including reaching, grasping and standing, have been investigated to develop our understanding of co-ordination.
Specific scientific achievements
The primary achievements in audition are (a) a set of experiments to demonstrate the role of neural, time-interval patterns in the perception of sound quality, and (b) the development of a computational model of hearing with a strobed temporal integration mechanism to convert repeating time-interval patterns into stabilised auditory images. The success of the resulting Auditory Image Model, AIM, has led to the development of a software package that simulates the activity patterns produced by sounds at various levels in the auditory system. A wide range of investigators are testing the package as a front-end for speech recognition systems, and as an explanatory tool for the role of time-interval information in the perception of sound quality. Research on auditory warnings has culminated in the production of a warning system for British Rail and the publication of European and International standards on hospital warnings.
In vision, a new system for precision ophthalmic tinting (the Intuitive Colorimeter System) has been developed from conception and patenting through to marketing, and it is now moving into general optometric practice. Preliminary clinical observations, an open trial, and a double-blind, placebo controlled, cross-over study have consistently demonstrated benefits in reducing eye¬strain and headaches.
The research in movement control has supported models of force and timing when subjects are explicitly required to control these aspects of movement. Empirical support has been obtained for a variety of visual cues contributing to interception skills. Under the heading of co-ordination it has been shown that grip force, used to stabilise an object in the hand, is modulated in anticipation of loads induced by arm kinematics. Finally, methods for investigating whole body stability have been developed.
Future plans for the next five years
In audition, the proposals for future research focus on the neural time-interval information produced by a sound and how the auditory system uses this and other information to separate signals from background noise in everyday life. The research will concentrate on two synthetic sounds which produce complex tone and noise perceptions that traditional auditory models fail to explain. The stimuli are matched pairs of damped and ramped sounds and iterated rippled noise. The perceptions are interpreted in terms of the time-interval patterns the sounds produce in the auditory image. The recent appointment of a second auditory scientist will enable us to expand our auditory research to source segregation. Although we experience sounds as separate sources, their acoustic waves are completely intermixed when they arrive at the ear; proposed research in this area will focus on how the individual sources are recovered by the listener.
The theory of visual stress suggests that precision tinting will also have therapeutic benefits in photosensitive epilepsy and migraine; the proposals for future work thus involve evaluating ophthalmic tinting in these disorders. The study of the mechanisms that underlie the beneficial effects will continue, with an emphasis on objective correlates of the subjective effects.
Proposals for future work on movement include evaluation of the control of force, timing and position in bimanual tasks with a view to synthesising a model of movement control from elemental component processes. Formal models for between-hand co-ordination within individuals will benefit from analyses of between individual co-ordination in rowing. Analyses of anticipatory processes involved in minimising the destabilising effects of voluntary movement will link studies of grasp and standing balance. The work will include studies of impaired function in the elderly and in neurological disorders of movement.
Implications for improving health, health care, and wealth creation
During the current reporting period, work in this programme has resulted in 6 patents, three of which have produced licence fees. In most cases, there is an industrial collaboration associated with the patent which increases the potential for wealth creation and widens the circle of dissemination of the underlying science. All of the patents have applications in healthcare.
In audition, it is anticipated that the auditory image model will be incorporated into a preprocessor for automatic speech recognition and eventually machine translation. The AIM software is distributed electronically to facilitate technology transfer and the dissemination of the science. The research on sound separation should lead to improvements in the operation of recognisers and hearing aids operating in noisy environments. Implementation of the new hospital warning standards should now begin to reduce preventable accidents in operating theatres and intensive care units.
In vision, further clinical trials of precision ophthalmic tinting may lead to an inexpensive treatment for light-sensitivity in migraine, a common component of a disabling disorder. The development of filters for use in the classroom may help to prevent reading difficulties in children. A search for objective correlates of visual discomfort may improve methods of assessment.
In movement control, investigations of standing balance have resulted in a patent that has lead to a commercial product, Swayweigh. This device will be useful to clinicians, therapists and patients in the diagnosis and remediation of balance problems including those resulting from stroke. Future research on balance after stroke and other neurological motor disorders should enable the development of fruitful new physical therapy approaches to rehabilitation. Research on two-hand co-ordination will suggest possible new dimensions for interaction with computers.
THE AUDITORY IMAGE (Allerhand, McKeown, R Patterson)
Introduction: Auditory Images and the Space of Auditory Perception
When an event occurs in the world around us, we experience an auditory image of the event, in the same way that we experience a visual image of the event. The auditory image reveals the pitch and loudness of the source and its sound quality, or timbre. These properties enable us to identify voices and musical instruments, and they tell us whether a speaker is angry or sad and whether a note is mellow or harsh. We have developed a theory of auditory image formation and a computational model to express the theory in concrete form. The theory distinguishes between the construction of auditory images by the peripheral auditory system and the processing of the images by the central auditory system. These are the two basic research themes in the report for this area. There are also two applied research themes: the application of auditory image theory to speech processing, and the design of auditory warning sounds. The report begins with a brief introduction to the concept of auditory images and the representation provided by the Auditory Image Model (AIM).
Auditory images are constructed in three stages, each of which creates a dimension of the space of auditory perception (Figure 1). The first two stages simulate the frequency analysis performed in the cochlea and the laterality analysis performed in
Figure 1. The space of auditory perception in the auditory image model.
Figure 2. The auditory image of the vowel /ae/, as in 'hat'.
the midbrain. Frequency and laterality are the vertical and horizontal dimensions of the plane in the centre of Figure 1. The activity generated by a compact sound source appears on a vertical line in this representation. The figure illustrates the separation of two sources, one 40 degrees to the right of the listener and containing energy in the mid-frequencies, the other 20 degrees to the left with energy at higher and lower frequencies.
In previous auditory models, this lateralised spectral analysis was assumed to represent peripheral auditory processing in its entirety, and all information concerning sound quality was assumed to be coded in the spectrum by relative intensity. The distinctive sound qualities we hear in music and speech, however, indicate that auditory image construction includes a sophisticated temporal integration mechanism and an analysis of the time intervals in the resulting neural activity pattern. It is as if the system maintained a large array of post-stimulus-time histograms behind the frequency-laterality plane, one for each frequency-laterality combination. Integration into the histogram is triggered by local maxima in the neural activity; the set of histograms activated by a point source form a vertical plane like those shown in Figure 1 behind the frequency-laterality plane. The structure that forms in the plane is the auditory image of the sound. When it is tonal, the histograms are regular and related as in the low-frequency channels of Figure 2 which shows a vowel presented in noise. When the sound is noisy, the plane contains irregular and unrelated histograms as in the upper channels of Figure 2 where the noise dominates.
In summary, auditory image theory suggests that peripheral auditory processing creates our space of auditory perception and our initial images of sounds. The auditory image represents our initial experience of the sound and is the basis of all subsequent auditory processing. Specifically, the theory provides a framework for understanding how the auditory system analyses acoustic events and how it characterises sources.
A. Auditory Image Construction
Patterson, Allerhand, Holdsworth, Akeroyd, Datta, Manson, Irino)
A1. Implementation:
The Defence Research Agency (Farnborough) awarded APU a grant to develop a reliable computational version of AIM, to tune it to calibration sounds, and to test its ability to measure stress in speech.
A1.1 AIM software development:
State-of-the-art software engineering principles were used to design a sophisticated platform and a user-friendly interface for AIM . This included (a) multiplex processing so that AIM can be applied to sounds of indefinite length, (b) an extensive display system that provides dynamic spectra and spectrograms as well as auditory images, (c) a source-code management system (SCCS) to facilitate maintenance of what is now more than 30,000 lines of code, (d) control programs to compile and build the model automatically, (e) routines to mail the package to collaborators electronically, and (f) routines to unpack and rebuild the package on a range of machines. In June 1991, AIM Release 5.0 was made available to collaborators (1.109).
A1.2 Tuning, testing and dissemination:
The model was tuned with click trains to set the temporal integration parameters, and with co-modulated tones to set the adaptation and suppression parameters (1.66). The system was tested with musical notes and diphthongs to show that the images of tonal sounds expand and contract horizontally as the pitch of the sound changes and the formants move up and down with changes in vowel quality. AIM was applied to pairs of concurrent vowels to illustrate how a pitch difference aids vowel separation in a time-interval model. It was also used to explain the sequence of perceptions that arise as the spectral density of a set of inharmonic sinusoids increases from one to five hundred (Profile Analysis). The adaptive thresholding mechanism and the image stabilisation mechanism were patented (1.119, 1.120). Invited papers were presented at conferences of physiologists (1.65), psychoacousticians (1.66), and speech scientists (1.85) to facilitate rapid dissemination of this new approach to hearing.
A2. Research:
The auditory images of everyday sounds reveal short-term temporal asymmetries; the formants of the vowel in Figure 2 are examples. Traditional auditory models treat these short-term asymmetries as phase changes which are removed entirely, or blurred, in the course of temporal integration. AIM preserves short-term asymmetries (1.79), and we believe they play an important role in the auditory analysis of sound quality. This led us to focus our research on a new class of sounds produced by applying asymmetric temporal modulation to continuous sounds. The strategy was to construct sounds with identical energy spectra that produced different auditory images and show that people hear differences in the sounds analogous to the differences in the images.
We used an exponential damping function as the modulator and reset the amplitude to its starting value every 25 ms to produce a sustained "damped" sound. The time-reversed version of this stimulus is referred to as a "ramped" sound. The continuous sound, or carrier, was either a sinusoid or a broadband noise, both of which have highly distinctive sound qualities. The damped and ramped versions produce major differences in what is heard, indicating (i) that temporal asymmetry is preserved in the auditory image even over short intervals, and (ii) that purely spectral models of hearing will have great difficulty explaining the timbre of everyday sounds.
A2.1 Damped/ramped tones:
Our first study revealed that the internal representations of damped and ramped sounds differ because the impulse response of the auditory filter is asymmetric and it interacts with the asymmetric envelopes of damped and ramped tones. The damped sound produces activity in a narrower range of auditory filters than the ramped sound, but listeners report that it is the ramped sound which has the stronger sinusoidal quality (1.18). The perceptual difference between damped and ramped sounds exists for a wide range of carrier frequencies and repetition rates. In these regions, the auditory image contains more time intervals at the period of the sinusoidal carrier for the ramped than the damped sound (1.19). It was concluded that it is time intervals at the appropriate period rather than a spectral peak at a specific frequency that determines whether a source will have a sinusoidal sound quality.
A2.2 Damped/ramped noise:
A wideband noise with a repeating ramped envelope sounds like a repeating noise; when the envelope is reversed, the resulting damped noise sounds like a shallow drum being struck with brushes! As with damped/ramped tones, the difference between damped and ramped noises exists for a wide range of stimulus conditions (1.63). In this case, however, the asymmetry does not arise in the cochlea. We have developed two mechanisms to explain the dramatic perceptual differences that we hear in damped and ramped noise. Both are based on the derivative of the envelope of the neural activity flowing from the cochlea; but whereas one is applied before temporal integration (1.82, 1.101), the other is applied during temporal integration. Experiments to determine which is the more appropriate site for the asymmetry are included in our future proposals.
In summary, these two new timbre discriminations demonstrate the importance of preserving short-term temporal asymmetry in our internal representations of sounds, and further that asymmetry arises both in the cochlea and at more central levels in the auditory system.
B. Auditory Image Processing
(Patterson, Allerhand, McKeown, Robinson, Akeroyd, Datta)
In the current reporting period, image processing was limited to research with a spiral representation of the auditory image (1.120).
B1. Octave Perception and the Spiral Auditory Image:
We performed two studies on octave perception, demonstrating (a) that many musical notes have non-integer octave ratings when compared with those of equivalent sinusoids (1.17), and (b) that the spiral form of the auditory image can explain the non-integer octave ratings (1.86). The studies also indicated that the auditory image is probably stabilised in the process of construction before, rather than after, pitch extraction. This hypothesis was supported by two series of experiments, one with short- duration musical notes which showed that the timbre of a note can be categorised more accurately than its octave or chroma (1.53, 1.54, 1.113,), and the other with short concurrent vowels which showed that the timbre of the dominant vowel is available from single cycle stimuli (1.52). As a result, work on image stabilisation displaced work on spiral image processing for the time being.
B2. Pattern Recognition and Speech Stress:
Following development of a computational form of strobed temporal integration, a pattern recognition mechanism was developed for the stabilised spiral output of AIM (1.64, 1.97, 1.98). It learns the forms of regularity that occur in a given data base and, once trained, can be used to identify the stressed syllables in continuous speech (1.78) or the degree of vocal agitation in the speech of pilots operating under emotional stress (1.94).
C. Application of AIM to Speech Sounds
(Patterson, Allerhand, Holdsworth, McKeown)
(Collaborators: Cutler, Norris, Fallside, Almeida, Schwartz, Hirahara, Anderson)
Speech recognition machines with traditional spectrographic preprocessors perform poorly in noisy environments, whereas human speech recognition is highly noise resistant. This led us to propose AIM as a preprocessor for speech recognition machines which, in turn, led to a series of collaborations in speech research.
C1. Auditory/Connectionist Techniques for Speech (ACTS):
The European Commission (ESPRIT BR) provided a grant to develop and test an auditory/ connectionist speech recogniser.
The data-rate problem: Traditional recognition systems use spectrographic representations of speech with data rates around 2 KiloBytes/second. The auditory image representation has an extra dimension and, as a result, its data rate is around 2 MegaBytes/second. Consequently, two modules were added to the AIM package so that the output of both the cochlea simulation and the auditory image could be time-averaged and down-sampled into a spectrographic form. This enabled us to interface AIM to (i) a Time Delay Neural-Net recogniser developed at APU, (ii) a Kohonen Net recogniser developed at INESC in Lisbon, and (iii) a recurrent error-propagation network developed in the Cambridge University Engineering Department (CUED) (1.85).
AIM and phoneme recognition: The most comprehensive recognition study was performed with the CUED recogniser; it involved training and testing the system with the complete TIMIT database (over 7000 sentences). It showed that phoneme recognition performance with an auditory/connectionist system was comparable to that from the best HMM recogniser available at the time (1.87). The other AIM/connectionist systems yielded similar results on smaller data sets. They are described in the annual project reports (1.110, 1.111, 1.112), along with about 20 papers on the development of the recognisers and their motivation.
C2. AIM as a Preprocessor for Speech:
In collaboration with the largest Japanese speech group at ATR in Kyoto, we developed an auditory/HMM phoneme recogniser (Patterson & Hirahara, 1989) that used a spectrographic reduction of the auditory image. It performed somewhat better than a traditional system when the size of the recogniser was constrained and the speech was noisy. The differences were significant but they were not large.
In collaboration with Armstrong Laboratories at Wright-Patterson Air Force Base in Ohio, we have set up an auditory/Kohonen-net recogniser to process Timit data. The system uses a spectrographic reduction of the neural activity pattern from AIM. It has so far demonstrated small but significant advantages over traditional recognisers in phoneme recognition (1.80) and speaker identification (1.84). Together, all of these studies make two important points:
(1) Auditory models are just as good as traditional preprocessors employed in speech recognition, even when the temporal fine-structure is removed to reduce the data rate, and even though they were not specifically tuned for speech sounds (1.85).
(2) To achieve the performance demostrated by human listeners, speech features will have to be extracted from a representation like the auditory image which contains fine-grained temporal information.
D. Auditory Warnings (Patterson, Datta, Milroy) (Collaborator: Lower)
Over the past decade, at the request of several government agencies (CAA, MOD, DTI, BR, BSI), we developed sets of auditory warning sounds to alert operators to life threatening situations in hospitals, aircraft and trains. A combination of auditory theory and ergonomics was employed to ensure that the warnings were, at one and the same time, more distinctive (therefore more memorable) and less startling (therefore less irritating) than existing commercial warnings. During the current reporting period, commercial systems were installed in civil helicopters flying to North Sea oil rigs and military helicopters sent to the Gulf war. The design principles were published with illustrations of their application in aircraft and hospitals (1.16, 1.89).
D1. Track-side Warnings:
Following our development of a system to alert track-side maintenance staff to approaching trains, British Rail Research requested that APU support the commercial development of the production version. Field measurements made at a test installation track beside the main line showed that some forms of the device did not meet the specified sound levels because the speaker was jammed in the housing (1.102, 1.105). Headsets were reviewed for use in extreme noise backgrounds (1.104), and finally, a procurement specification and quality control procedure were written (1.103). This work is now complete.
D2. Civil Aircraft Warnings:
Following two airline disasters, the CAA asked us to design a warning to aid passengers evacuating a smoke-filled aircraft. Tests with warnings mounted near exits demonstrated that they were readily localisable (1.106), and an evacuation test in a BAC 111 showed that the system did improve the speed of evacuation from the smoke-filled fuselage. This work is now complete.
D3. Hospital Warnings:
We provided support to DTI during the development of hospital warning standards by the BSI in the UK, by CEN in Europe, and by the ISO internationally. Recently, these draft standards received majority votes, and so they will become mandatory standards in the near future.
D4. Military Aircraft Warnings:
Recently, DRA requested that APU support a multi-national effort to develop localisable auditory warnings to tell the pilot whether there is an aircraft problem or an external threat. Our initial task was to extend the frequency range of the existing helicopter warnings to 12 kHz without changing their character. Three methods were developed for extending the frequency range; each method proved successful with at least some of the existing sounds (1.108). This is a continuing project as described in section D4.
FUTURE RESEARCH AND TECHNOLOGY TRANSFER
A. Auditory Image Construction (Patterson, Carlyon, Allerhand, Akeroyd, Datta, Irino) (Collaborators: Giguerre)
A1. Implementation:
Two recent developments lead us to believe that is time to produce a professional version of AIM, both as a means of disseminating our basic research and as a means of facilitating the transfer of AIM technology to industry. The first development involves an American programme to demonstrate the value of analogue VLSI chips through development of a 'silicon retina' and a 'silicon cochlea'. Through our collaboration at Armstrong Laboratories, AIM has been chosen as the model for the 'silicon cochlea'. If successful, the chip will greatly accelerate the use of auditory models in hearing, speech and music research. The chip is also intended to support commercial development of verbal computer interfaces. The second development occurred at APU. In August 1993, we began regular distribution of AIM to our collaborators by anonymous ftp. APU records the traffic into ftp accounts; to our surprise, we found that 125 sites acquired the full AIM package in the eight months following its installation. Although we do not imagine that all 125 investigators are using AIM regularly, this take-up rate does indicate that there is demand for an auditory modelling package. This section presents our proposals for a professional release of AIM.
A1.1 A physiological cochlea for AIM:
A state-of-the-art, physiological cochlea simulation has been developed in the Cambridge University Engineering Department. They used the AIM software platform and replaced our linear, feed-forward cochlea with a non-linear, active-feedback cochlea. The physiological cochlea can simulate auditory distortion products and cochlear echoes and, as a result, it can be used to support research on hearing loss at the cochlear level. With financial assistance from the Cambridge Psychology Department, we are installing the physiological cochlea in parallel with our functional cochlea in the AIM software package. We propose to create demonstrations to illustrate the conditions under which one or the other of the cochlea simulations is preferred. The physiological cochlea will enable us to collaborate on the modelling of hearing disorders with the hearing group in the Psychology Department.
A1.2 A correlogram module for AIM:
There are several auditory models that stabilise time-interval patterns from the cochlea using multi-channel autocorrelation rather than strobed temporal integration. The resulting correlograms are similar to auditory images but more symmetric. We have derived the underlying mathematical relationship between the correlogram and auditory image models (1.79). We have written a correlogram module that is compatible with the AIM software, and we propose to integrate it into the package to enable direct comparison of the correlograms with auditory images as representations of what we hear. When completed, the AIM package will be able to simulate most of the models currently used for auditory perception -- including traditional spectral models.
A1.3 An X-Windows interface for AIM:
The current command-line interface to AIM provides fast, flexible control of the model. Nevertheless, it requires some learning on the part of the user. Software engineering research indicates that an X-Windows, menu-driven interface is essential if AIM is to gain user acceptance beyond the Unix-programmer community. There are now excellent tools available for building X-Windows interfaces, and a pilot project to provide such an interface for our image review system was highly successful. We also propose to develop support facilities for AIM in the form of on-line documentation, demonstration scripts, a library of example sounds, and a package of tools to facilitate auditory image processing. The on-line manual will be compatible with the widely used UNIX man system.
A1.4 Auditory image video facilities:
Using relatively crude techniques, we have made videos of the auditory images produced by dynamic sounds and found that they are an invaluable tool for explaining the auditory image concept. The videos lead us to believe that real-time auditory-image displays will also prove valuable for remedial speech training with children and as teaching aids for musicians. Indeed, a plug-in peripheral for televisions, with a silicon cochlea and real-time image stabilisation, would enable one to watch auditory images of music and speech while listening to the sound. While this may seem a frivolous illustration, it is worth noting that it is the entertainment market that originally brought down the cost of video machines. In any event, we believe that auditory image videos will play an important role in expanding the use of auditory models and the public understanding of science. To facilitate this development, we propose to purchase a video system that will play auditory-image bitmaps to a video recorder at the appropriate rate. Previous methods of producing videos have resulted in low-resolution, jerky images produced with great labour.
A2. Research:
The perception of damped and ramped sounds will be used to refine the mechanisms used in the construction of auditory images.
A2.1 Damped/ramped tones:
A ramped tone generates carrier activity across a wider range of frequencies than a damped tone, and the size of the difference observed in the auditory image is affected by the degree of compression in preceding stages. In the functional version of AIM, the compression function is logarithmic; in the physiological version, it is linear at low and high levels and compressive at moderate levels. We propose to perform two sets of experiments to determine the appropriate form of the compression function for a perceptual model of hearing.
Matching the sinusoidal component of damped and ramped tones: As previously (see 1.18, 1.19), listeners will be presented a ramped and a damped tone in random order and asked to choose the one with the stronger sinusoidal component. The half-lives will be varied to identify the region in which this is a difficult decision. It is assumed that these matched pairs generate roughly the same number of carrier periods in the biolgical auditory image, and so we will adjust the compression in AIM to achieve a carrier-period match in the simulated auditory image. The experiment will be replicated at different intensity levels, since the physiological cochlea simulation is level dependent.
Masking by pairs of damped and ramped tones: The shape of the auditory filter can be measured by positioning two masking tones near a narrow band of noise and measuring threshold for the noise signal as a function of the frequency difference between the masking tones. We propose to employ this technique (a) with a damped noise and damped tonal maskers and (b) with a ramped noise with ramped tonal maskers. If the perceptual difference between damped and ramped tones arises in the cochlea as hypothesised by Patterson (1.18), the filter shape derived with the ramped stimuli will be wider than that derived with the damped stimuli. Again, the size of the difference depends on the degree of compression, and the AIM compression value will be adjusted in accordance with difference between the two filter shapes.
A2.2 Damped/ramped noise:
Enhancement of a frozen segment of noise in a jittered damped noise. The strobed temporal integration mechanism in AIM has the unique property of increasing the signal-to-noise ratio of sound components that occur a fixed interval from a strobe pulse repeatedly. We will attempt to demonstrate that such components are enhanced in the auditory system. A jittered damped envelope will be constructed in which the envelope peaks occur every 30 ms, on average, but with individual temporal displacements of 0, 5 or 10 ms in either direction. The envelope will be applied to a random noise into which a 5-ms segment of frozen noise will be inserted every 30 ms, on average. In one case, the spacing between occurrences of the frozen segment will be fixed at 30 ms, independent of the position of the envelope peak; in the other case, the frozen segment will be shifted to be a fixed interval from the envelope peak. We will measure the detectability of the frozen noise segment as a function of its position relative to the peak. In AIM, the envelope peaks will drive the strobe units and so we would expect to find that the frozen noise is more detectable when it occurs a fixed interval from the envelope peak.
A2.3 The spectral spread of strobe pulses:
In AIM, the strobe units that initiate temporal integration operate independently; that is, they search for local maxima only in their own channel. Our auditory images of periodic sounds are extremely stable, which suggests that a large local maximum in one channel may entrain strobe units in nearby channels with relatively weak local maxima. We propose to perform a masking release experiment to determine whether there is cross-channel strobe interaction. A stream of tone pips will be presented in noise and adjusted to be near threshold. The mean interval between tone pips will be 30 ms but the precise position will be shifted a random amount up to 10 ms in either direction. We will then introduce streams of clicks in channels above and below the tone-pip channel; the mean inter-click interval will be 30 ms. In one condition the clicks will be synchronised to the tone pips; in other conditions the click stream will be either random or perfectly regular. If threshold is lowest when the clicks are synchronised to the tone pips in the synchronised click condition, it will indicate that synchronised strobe activity in a range of channels entrains strobe units in adjoining regions when the strobe stimulus in that region is weak.
B. Auditory Image Processing (Patterson, Carlyon, Allerhand, Akeroyd, Datta, Handel) (Collaborators: Yost, Fay, Feth)
B1. Octave Perception and the Spiral Auditory Image: The spiral representation of octaves will be required to explain the form of musical scales, and it offers a solution to the problem of octave errors in speech recognition. Currently, however, it is not a priority.
B2. Pattern Recognition and Sound Quality Discrimination:
The auditory image measures developed at APU (pitch strength and loudness (1.78)), and the correlogram measures developed at Loughborough (concurrent vowel heuristic) will be developed into a set of metrics for sound quality discrimination. The purpose is twofold: 1) to establish a metric for the degree to which a given auditory model can explain existing discrimination data concerning pitch, damped and ramped sounds and concurrent vowels, and 2) to develop a comparison metric for evaluating different auditory models (e.g. AIM and the correlogram model). Together these metrics will also enable us to determine when a proposed improvement, like the transmission line filterbank or the Meddis transduction module, makes a quantitative difference to our ability to explain discrimination data.
B3. Time-Interval Patterns and Sound Quality:
At the start of his famous book, Helmholtz (1885) states that the most prominent distinction in auditory perception is that between tones and noises, and that the first job of auditory scientists is to explain this distinction. Broadly speaking, tones are produced by periodic sounds and noises by aperiodic sounds; but little is known about just how regular a wave must be for us to hear a tone, or how irregular a wave must be for us to hear the characteristic shshsh of noise. A pair of damped and ramped noises with the same half-life have the same degree of irregularity in physical terms, but one produces a shshsh perception and the other does not. Our work with auditory images suggests that the distinction between tonal and noisy perceptions is determined by the degree of regularity in restricted regions of the auditory image as opposed to regularity in the acoustic wave per se, and that this distinction sets up a figure/ground relationship in the auditory image. In the next reporting period, we propose to investigate these hypotheses and attempt to model them in terms of regularity in the auditory image.
B3.1 The perception of iterated rippled noise (IRN):
Rippled noise is constructed from a random noise by delaying a copy of the random noise and adding it back to the original. The delay-and-add process introduces ripples into the spectrum, with peaks at multiples of the reciprocal of the delay and valleys midway between them. Rippled noise sounds like a pair of concurrent sources, a weak tone with a prominent broadband noise. When the delay-and-add process is repeated, or iterated, the tonal component of the perception grows stronger and the noise component grows weaker; by about 10 iterations, the noise component of IRN is barely noticeable. When the delay is long (say 16 ms), the spectral peaks are closely packed (every 62.5 Hz) and, in the region above about ten times the peak spacing, the auditory spectrum of the IRN is quite similar to that of random noise. Thus, a spectral model of hearing would suggest that, if random and iterated noises are highpass filtered and equated for energy, they should not be discriminable. Nevertheless, they are perfectly discriminable; one has the shshsh of noise and the other sounds like a buzzy musical note.
The timbre contrast between random and iterated noises led us to suspect that IRN would be more detectable in random noise than in IRN, and that random noise would be more detectable in iterated noise than in random noise. A short experiment was performed with an IRN having 256 iterations. When the masker was a random noise, the IRN signal was 4 dB easier to detect than a random noise signal; when the masker was an IRN, the IRN signal became 6 dB harder to detect while the random noise became 14 dB easier to detect. Thus, the availability of a sound-quality cue based on temporal regularity leads to a 20-dB interaction in masking threshold where a spectral model would predict no difference whatsoever! We propose to measure the interaction as a function of the position of the highpass filter. If the pitch is based on spectral peaks, then performance will improve as the delay decreases. If the discrimination is based on fine-grained time-interval information, then the interaction should decrease as the highpass cutoff increases, since auditory phase locking decreases at high frequencies.
B3.2 Auditory figure/ground:
The auditory images of tonal sounds are static, complex, 3-D figures with regular interiors (Figure 2); the auditory images of noises are scintillating, featureless, irregular fields. When they occur together, there is a strong figure/ground separation between the components of the image dominated by the tonal sound vs the noise. We believe that figure/ground separation is probably the first stage of image processing in the auditory system and that it is based on temporal regularity in the auditory image. We propose to develop measures of regularity based on the auditory images of IRN; there are both regular and irregular regions in the images of IRN and the size of the regular region grows with the number of iterations. We will begin with measures that locate concentrations of regularity in individual channels and peak alignment across channels. In this way we hope to develop a method of isolating auditory figures in images and separating them from the irregular regions associated with background noise.
C. Application of AIM to Speech Sounds (Patterson, Allerhand, Datta) (Collaborators: Anderson, Kawahara)
C1. Auditory/Connectionist Techniques for Speech (ACTS):
The Esprit BR project imposed an excessive burden of administration on the auditory group at APU. This drew us away from our core research, and so when the Esprit project ended, we withdrew from the management of large speech projects.
C2. AIM as a Speech Preprocessor:
Our speech recognition work will continue through our international collaborations. The collaboration with Armstrong Laboratories on phoneme recognition will be extended to word recognition, where we will attempt to show that an AIM/HMM system supports word recognition to the same level as systems with traditional preprocessors. We are also planning to develop a binaural version of AIM at Armstrong Labs with the assistance of Bochem University in Germany. The purpose of this system is to segregate a speech source from diffuse background noise binaurally before reducing the neural activity pattern to a spectrographic form. This system has the potential to show a dramatic improvement for speech recognition in noise. It is also through Armstrong Labs that we collaborate on the 'silicon cochlea' programme.
The collaboration with ATR in Kyoto will continue through our participation in a programme to develop a biological framework for speech perception and perception. Our contribution involves research on source separation through auditory figure definition.
C3. Speech Cleaning with AIM:
There is a new algorithm for resynthesising sound from correlograms and auditory images. This suggests that it may one day be feasible to 'clean' speech for recognition systems, hearing aids and communication systems by (a) isolating the voiced parts of speech in auditory images on the basis of their internal regularity, (b) separating the voiced parts of the speech from background noise through their figure/ground relationship, (c) resynthesising the voiced speech, and (d) recombining it with a reduced background component. We will look for an opportunity to assemble a software prototype of such a cleaning system in collaboration with one of the resynthesis groups.
D. Auditory Warnings from AIM (Patterson, Datta)
D4. Military Aircraft Warnings:
DRA has a requirement for new warning sounds to signal threats to an aircraft from the ground or the air. The individual threat warnings should be separately identifiable. At the same time, they should also have a distinctive general form to indicate that they are threat warnings rather than flight-systems warnings. We propose to use damped and ramped sounds with complex carriers as building blocks for the new warning sounds, and to use AIM to as the design platform. Individual cycles of damped and ramped sounds will be assembled into groups with distinctive, partially random, inter-pulse intervals to add a degree of sharpness and buzz to the sounds. Then the groups will be combined with partially random inter-group intervals to give sounds with a roughness and sharpness that make them urgent in a way that has not been heard before.
CONCURRENT SOUND SEGREGATION (Carlyon)
Introduction
This part of the report describes the research of Dr. R. Carlyon, who joined the Unit in April 1994. The discussion of his previous work is provided as a background to his proposals for future research.
Background
The task of listening to one sound, such as a voice, in the presence of a competing sound, such as a second voice, is one of the most demanding faced by the auditory system. Not only does the composite waveform of the two sources have to be analysed into its constituent frequency components, and the changes in those components tracked over time, but the system also has to determine which components belong to which source. This problem is exacerbated by the fact that, in the case of speech, the two voices will usually have frequency spectra which overlap, and will share similar temporal characteristics. Much of Carlyon's research has investigated how we compare information in different frequency regions in order to separate concurrent sounds perceptually. Typical tasks have required the listener to detect differences in fundamental frequency (F0, the physical analogue of pitch) between two groups of frequency components, or to try to detect differences in the pattern of frequency modulation (FM) imposed on two individual components. Because many of his recent findings form the impetus for future experiments, they will be described in later sections. However, it is worth making two general points regarding those experiments. First, they used forced-choice tasks, requiring listeners to discriminate between a stimulus that would normally correspond to a single source, and one which would normally arise from two sources. This method has several advantages over other procedures, such as those requiring listeners to rate stimuli according to their degree of "perceptual fusion": one advantage is that it allows a rigorous, quantitative test of different auditory models, for example by the application of signal detection theory. Second, care was taken to ensure that listeners were definitely performing across-frequency comparisons: the components to be compared were always well-separated in frequency, and any within-channel interactions were masked by noise. Across-frequency comparisons are useful when there are regions of the combined spectrum of two sounds which contain energy from only one source, and where the listener needs to assign those regions to the appropriate sound -- for example, where the formants of two vowels from different speakers are interleaved across the frequency spectrum.
Several of the new experiments proposed are motivated by the fact that, in many situations, formants are not simply interleaved: instead, there are frequency regions which contain energy from both sources. Therefore, in addition to extending Carlyon's earlier work on across-frequency processing, we will investigate the within-channel cues that listeners use to segregate concurrent sounds. In this case, however, it is not appropriate to require listeners to discriminate between a sound consisting of two sources (e.g., two F0s), and one consisting of a single source. This is because some within-channel cues, such as beats, can be detected using a minimum of processing, without necessarily being useful for concurrent sound segregation. The experiments on within-channel cues will therefore require listeners to segregate two components, or groups of components, that are very close in frequency, and then to compare one of these to a third sound, which is separated from the other two in either frequency or time.
FUTURE PROPOSALS
A. Extracting Two Periodicities in the Same Frequency Region (Carlyon)
A1. Background and Pilot Experiments.
One way of segregating two sounds that occupy the same frequency region is by differences in their F0s. Although listeners can exploit the cue using only across-channel mechanisms (Carlyon et al., 1992) , and its use in concurrent sound segregation is well established for stimuli where both within- and across-channel cues are available (Scheffers, 1983; Assmann & Summerfield, 1990) , the potential usefulness of within-channel cues has received less attention. We will address this issue using stimuli consisting of two harmonic complexes, each bandpass filtered identically but with slightly different F0s. In pilot experiments, listeners have been presented with a 500-ms harmonic complex, consisting of a large number of harmonics of 210 Hz passed through a bandpass filter. After the first 150 ms of this sound, a second complex, with the same level as the first and passed through an identical filter, was added; its F0 was either 175 or 252 Hz and its duration was 200 ms. The perceptual results of this simple manipulation were striking: when the filter cutoffs were set to 20 and 1420 Hz, so that the components of each sound were resolvable by the auditory periphery, listeners heard the second "target" sound as having a clear pitch. However, when the filter was set to 3900-5400 Hz, so that the components were unresolved, the target sounded distinctly aperiodic: a common description was that it "crackled". A roughly similar percept occurred when the duration of the 500-ms "masker" was reduced to 200 ms, so that the two complexes were turned on and off together.
Our perceptual observations have, to some extent, been reinforced by the results of a two-interval discrimination task, in which the target's F0 was lower than that of the 500-ms "masker" in one interval, and higher (by the same amount) in the other: listeners' ability to identify the interval with the higher F0 was measured as a function of the F0 difference (∆F0) between the two targets. For the lower filter setting (resolved components), performance improved monotonically with increasing ∆F0, whereas for the higher setting (unresolved components), performance remained at chance even for the largest ∆F0 (32%) studied.
A2. Can We Extract the Periodicities of Two Overlapping Groups of Unresolved Components?
The above findings provide some tentative evidence that listeners are very bad at extracting the periodicities of two groups of unresolved components which occupy the same frequency region. The proposed experiments will test this preliminary conclusion, will obtain perceptual measures of the effect, and will explore the similarities and differences between these perceptual reports and the results of discrimination experiments. It seems likely that at least some aspects of the phenomenon can be explained using R. Patterson's auditory image model ("AIM"; see Patterson's section of this programme), whose triggering mechanism needs a regular, periodic input in order to form a stable image: the existence of more than one trigger per target period will prevent the target from being represented in a stable form. Our general approach will be to present our stimuli to a number of auditory models, in an attempt to identify the most convincing and parsimonious explanation for our results.
One reason for caution in accepting our preliminary conclusion is that when, in an additional condition of our discrimination experiment, the masker and target were turned on and off synchronously, performance for the unresolved components was better than in the "asynchronous" condition, in which the masker started before the target and ended after it. One explanation for this result is that listeners can extract the two envelope periodicities, but that the portion of the masker preceding and/or following the target (the "forward" and "backward" fringes) somehow reduced performance in the asynchronous condition. At present, we consider this second explanation unlikely, as it does not account for the fact that gating the masker asynchronously did not degrade performance for groups of resolved components. An alternative explanation, consistent with our preliminary interpretation, is that gating the masker and target synchronously caused them to become perceptually fused, and allowed listeners to base discrimination on the average F0 of the masker and target, which differed between the two intervals of each trial. We will perform two experiments designed to tease apart these two classes of explanation.
The first experiment tests one specific explanation, that of "backward recognition masking" (Massaro, 1975) , according to which the backward fringe of the masker interferes with the processing of the target's F0. We will determine whether the effects of masker gating are mainly due to the backward or to the forward fringe of the masker. Pilot investigations tentatively suggest that the forward fringe is the more important, a finding which, if confirmed, is inconsistent with backward recognition masking; "forward recognition masking" is a much weaker effect. Second, we will perform an experiment using asynchronous maskers in which the target will be presented at the same time as a third, "reference" group of components, and the task will be to say whether the target and reference groups are presented on the same, or different, F0s. Two conditions will be included: either the masker and target will be filtered from 3900-5400 Hz and the reference from 20-1420 Hz, or masker and target will be filtered from 20-1420 Hz and the reference from 3900-5400 Hz. Thus, in both conditions, the listener has to compare the F0s of two groups in the frequency regions 20-1420 Hz and 3900-5400 Hz; the conditions differ in which region contains the masker. If listeners can extract the F0s of two groups of resolved components in the same frequency region, but not those of two unresolved groups, then they should perform above chance only when the masker is in the 20-1420 Hz region. However, if the results of our sequential discrimination task were due to some effect of the forward or backward masker fringes on the memory for the target's F0, then listeners should perform above chance in both conditions of the simultaneous task.
A3. Resolvability or Absolute Frequency?
So far, we have discussed the effect of frequency region on our results in terms of the resolvability of components, rather than in terms of absolute frequency (such as might occur, for example, due to phase locking falling off at high frequencies). Our sequential discrimination paradigm will be used to distinguish between these two interpretations, by orthogonally varying F0 and frequency region.
A4. Refining Perceptual Measures of the "Crackle" Percept: Following on from our perceptual experiments with informal listening, we aim to provide a more quantitative measure of the "crackle" percept, so that its dependence on parameters such as frequency region, F0, and masker gating can be compared with the results of our discrimination experiments. One approach will be to ask listeners to rate the perception of the target on a continuous scale, ranging from "definitely a periodic sound" to "definitely a crackle". To accomplish this, we need to generate a sound which can played to listeners as an example of what is meant by a "crackle", and, for reasons of circularity, this demonstration sound cannot be any of the test sounds. We have informally observed that a crackle percept can be produced by a random train of equal-amplitude spikes, where the duration of each spike is 0.1 ms and the probability of a spike occurring in any 0.1-ms time bin is about 5%. We will describe in more detail the percepts generated by random spike trains in section D; here we simply point out that such sounds have a wider bandwidth than our mixture of two groups of unresolved components, but can capture the same "crackle" quality. In order to equate the audible bandwidths of the spike trains and harmonic complexes, both sounds will be presented in a background of continuous bandstop noise. A second approach will be to require listeners to match mixtures of one periodic complex and one spike train, to a mixture of two periodic sounds: again, both sounds will be presented in bandstop noise, and listeners will adjust the relative amplitudes of the periodic and aperiodic parts of the former mixture to sound "equally periodic" as the latter.
A5. Further Investigations:
Our new perceptual measures will be used, in tandem with our discrimination task, to investigate two further aspects of the phenomenon which have so far been observed only in informal listening. One of these is that, although a 200-ms target/masker mixture sounds aperiodic, increasing its duration to several seconds allows the listener to hear the two underlying periodicities. We will investigate the time course of the effect, which, if substantiated, would require a revision of existing auditory models, including R. Patterson's AIM and those based on correlograms (Slaney & Lyon, 1990; Meddis & Hewitt, 1991) . Second, when the masker contains a forward and a backward fringe, listeners report that they can hear its pitch continue throughout the (aperiodic sounding) target. However, it is not clear whether this information is genuinely preserved in the auditory periphery (such as might occur in the AIM if it continued to trigger on the peaks in the masker envelope), or whether it is merely an example of central perceptual processes "filling in" information which is no longer present (Ciocca & Bregman, 1987) . This will be determined by imposing frequency modulation (FM) on the portion of the masker preceding the target, but keeping that part of it which is synchronous with the target unmodulated. If information on the masker's F0 remains available during the target, then listeners should hear the modulation cease: if, however, they are simply filling in what the masker "ought to" sound like, then they should hear the modulation continue.
B. Within-Channel Processing of FM Incoherence (Carlyon)
When the fundamental frequency (F0) of a periodic sound changes, all the harmonics of that F0 change frequency coherently (in the same direction, at the same time). It has been suggested that listeners use this FM coherence to group together the commonly varying components of one sound, and to separate them perceptually from components of other sounds, which may be changing frequency differently. However, Carlyon (1991; 1994b) has shown that, for resolved frequency components, listeners cannot discriminate between coherent and incoherent FM, except by virtue of the fact that incoherently modulating one component of a harmonic complex causes it to become mistuned from the rest. He concluded that there is no across-frequency mechanism specific to the processing of FM incoherence. In those experiments, within-channel cues were eliminated, for fear that the subject would make the discrimination on the basis of some trivial aspect of the stimulus. In our new task, listeners will be required to compare the frequency of one component of a closely-spaced harmonic complex, modulated coherently or incoherently with the rest, to that of a subsequent tone presented in isolation.
In one condition, the first stimulus presented on each trial will consist of harmonics 14 through 26 of a 100-Hz F0, with the 20th harmonic (the "target") mistuned upwards on half the trials and downwards on the other half. All harmonics will be frequency modulated coherently by +/- 2.5% of their carrier frequencies, and at a rate of 2.5 Hz. The second half of each trial will consist of a single component, modulated at the same depth and rate as the target, but with a frequency which differs from it by plus or minus ∆f Hz . In both conditions, the overall FM phase will be randomised from trial to trial, but the second, isolated, tone will always be modulated coherently with the target. The listener's task will be to decide whether the carrier frequency of the second tone is higher or lower than that of the target; the overall frequency will be randomised between trials in order to force the listener to make this comparison, rather than perform the task on the absolute frequency of the isolated tone. Existing data (Hartmann et al., 1990; Moore & Ohgushi, 1993) indicate that performance on the task will increase with increasing mistuning, and we will measure psychometric functions (d' vs. mistuning) to verify this, for a range of values of ∆f.
In the second condition, the mistuning will be produced by modulating the target incoherently with the other harmonics. The FM incoherence will cause the target to become mistuned by an amount which varies sinusoidally throughout the stimulus, and which increases monotonically as the modulator phase disparity between the target and other components is raised from 0˚ to 180˚. We will exploit the fact that it is possible to calculate the "equivalent maximum mistuning" produced by a given modulator phase disparity, and will measure sensitivity for a set of disparities which correspond to the mistunings used in the first condition. If FM incoherence provides a cue to segregation over and above the mistuning that it causes, then we might expect performance to be better in the second condition than in the first. If, however, FM incoherence is only effective because of the resulting mistuning, we would expect performance to be comparable, and for sensitivity to vary with "equivalent" mistuning in the same way in the two conditions -- in other words, the slopes of the functions in the two conditions should be similar (Carlyon, 1991) .
C. How Many F0-encoding Mechanisms are there? (Carlyon)
Traditional theory assumes that the mechanism used to estimate the F0 of low-numbered harmonics (which are resolved by the auditory system into separate frequency channels) is completely different from that used for the F0 of high-numbered harmonics, which are unresolved. In contrast, recent, more parsimonious, theories (including both AIM and the "correlogram" models) have proposed a common mechanism for deriving the F0s from both resolved and unresolved harmonics. Recently, Carlyon and Shackleton (1994) have presented evidence against this "single-mechanism" class of theory: although listeners are very good at comparing the F0s of two groups of resolved harmonics or two groups of unresolved harmonics, they are very bad at making the comparison when one group is resolved and the other unresolved. Those results are consistent with the idea that, when two groups of harmonics differ in "resolvability", they are processed by separate mechanisms, the outputs of which must be "translated" into a common format, and that this translation stage introduces a significant amount of inaccuracy into the comparison. This conclusion was backed up by a detection theory analysis, which compared sensitivity on the across-frequency ∆F0 task to that on sequential F0 differences imposed on the individual groups used in any one comparison. The purpose of the sequential task was to provide an estimate of the accuracy with which each group's F0 was encoded by the auditory system. The analysis showed that only when the two groups differed in resolvability was performance on the simultaneous across-frequency comparison worse than predicted from the "encoding accuracy" of individual groups.
Carlyon and Shackleton's findings have important implications for theories of pitch perception and of concurrent sound segregation, but are complicated by two factors, both of which can be overcome fairly straightforwardly. First, when listeners are asked to detect a sequential difference in the F0 of a group of resolved harmonics, they might perform the task simply by doing a frequency discrimination on the individual harmonics within that group: if so, they could "bypass" the F0-extraction mechanism, and performance on the sequential task would over-estimate the accuracy with which the F0 of resolved harmonics is estimated. This will be overcome by presenting only a subset of the harmonics in each group on any one interval, both for the sequential and the simultaneous comparisons. As the harmonics will differ on the two halves of each trial, listeners will be unable to perform the sequential frequency discrimination on individual harmonics, but will be forced to detect a difference in F0 per se. Second, when two groups of unresolved harmonics are present, there is evidence that listeners detect ∆F0s by virtue of the "pitch pulse asynchronies (PPAs)" that inevitably arise: the peaks in the waveform of the group with the lower F0 lag progressively further and further behind those of the upper group (Carlyon, 1994a; Carlyon & Shackleton, 1994) . This cue is not available when one of the groups is resolved (Carlyon, 1994a) , and is not available in a sequential comparison. The result is that, for unresolved harmonics, performance on the sequential task substantially under-estimates that on the simultaneous, across-frequency comparison, due to the existence of an additional cue in the latter task. Fortunately, the PPAs that result from a ∆F0 are small immediately after stimulus onset, and so they can be reduced by shortening the duration of the stimulus from 400 ms to 50 ms: if necessary, performance can then be increased to a measurable level by repeating the same 50 ms segment several times, without increasing the PPAs (cf. Assmann & Summerfield, 1994) . These two manipulations, combined with Carlyon and Shackleton's findings, should provide strong evidence as to whether the F0s of resolved and unresolved harmonics are encoded by the same, or by different, mechanisms.
D. Perception of Random Spike Trains (Carlyon, Patterson)
We have generated a continuum of sounds consisting of consecutive 0.1-ms time bins, in which the probability of any bin being filled with a spike (as opposed to silence) ranges along the continuum from 0.1% to 50%. At low probabilities (<1%, average interval between spikes > 9.9 ms), the listener hears a train of individual clicks, which turns into a crackle percept as the probability increases further. At still higher probabilities, one hears a "shhh" similar to Gaussian noise. This continuum of percepts has two implications for auditory models such as that proposed by Patterson. (i) Most temporally based models do a good job of distinguishing between periodic and aperiodic sounds; however, the "crackle" and the "shhh" are both aperiodic, and have similar bandwidths. Can the models reflect the difference in perception, and, if not, what modifications need to be proposed? (ii) The continuum reflects the operation of two types of temporal integration, with different time constants, operating on similar stimuli. One, with the longer time constant, relates to the transition between "multiple clicks" and a single crackle sound. The second, at the transition between the crackle and "shhh" percepts, has a shorter time constant. Our hypothesis is that, when several spikes occur within this short time window, the bimodal distribution of amplitudes in the spike train is converted to an approximately Gaussian distribution at the output of the window by the random nature of the spike generation process. The stimulus will allow us to measure the time constant of both types of integration over a wide range of supra-threshold levels, and will provide a strong test of the temporal integration parameters of existing auditory models. In particular, we will test the idea that the shorter time constant corresponds to the impulse response of auditory filters responding to the sound, by playing bursts of the sound in continuous notched noise. If so, then, as the centre frequency of the notch is increased, the shorter impulse responses of high-frequency auditory filters should cause the transition between the crackle and "shhh" percepts to occur at progressively higher spike probabilities.
VISUAL DISCOMFORT
(Wilkins)
A. A General Unified Theory
A general and unified theory of the neurological mechanisms underlying visual discomfort has been advanced in a book shortly to be published by Oxford University Press (1.1). In brief, the theory draws upon similarities between the visual stimulation that induces seizures in patients with photosensitive epilepsy and the visual stimulation that, in others, evokes perceptual distortion, eye-strain and headaches. According to the theory, the perceptual distortions are due to a localised and constrained hyperexcitability of neurones in the visual cortex. The distortions are associated with headache susceptibility in many different ways. They are also associated with susceptibility to one of the common cold viruses (respiratory cyncitial virus) which predisposes to asthma, with which migraine is statistically associated (1.20). The visual stimulation responsible for perceptual distortion is uncomfortable and interferes with the perception of other weaker stimuli. The extent of this interference is greater in people who suffer migraine (1.99).
The theory of visual discomfort has been applied to reading in two review chapters (1.71, 1.72), particularly in relation to the spatial aspects of text (1.88) and the complaints that surround the use of visual display terminals (1.33). The theory also provides a parsimonious interpretation of the therapeutic effects of ophthalmic tinting, suggesting a potential for therapy not only in reducing the perceptual distortion suffered by children with reading difficulty, but in preventing photophobia in migraine and reducing liability to photosensitive seizures.
The theory of visual discomfort has been applied to lighting in three review papers (1.34, 1.70, 1.90). Patients with agoraphobia have a sensitivity to light and, under double-blind conditions, their heart rate is affected by the imperceptible 100-per-second modulation of light from fluorescent lamps (1.12). This modulation has been measured for most types of commercially available fluorescent lamps (1.36), and the data thus obtained has been used to design an ophthalmic tint that reduces the modulation with a minimal effect on colour perception, and with a reasonably cosmetic colour appearance (1.42). This tint has been patented (1.122) and is marketed by Cambridge Optical Group Ltd. as the "Comfort 41". Large individual differences in preference for illuminant chromaticity have been demonstrated (1.92).
B. Precision Ophthalmic Tinting
Certain individuals, particularly those with a family history of migraine, report perceptual distortions of text when they read, sometimes sufficient to impede or impair reading (1.39). These individuals usually report that the perceptual distortions abate when the text has a particular colour, different for each individual (1.41). Individuals with a family history of migraine tend to report discomfort when the text has a reddish hue, a tendency that age and sex matched controls do not show (1.5). A simple but novel device (Intuitive Colorimeter) has been developed and patented (1.121). It illuminates a page of text with coloured light in such a way that the hue and saturation of the colour can be varied separately without an associated change in brightness (1.40). Small changes in colour can be evaluated whilst the eyes are colour-adapted. Colour constancy mechanisms do not complicate the judgement because all surfaces have even spectral reflectance. The analysis of the colour changes has involved a novel application of statistical methods which allow for circular rather than linear variables (1.4).
The Colorimeter provides a quick and accurate method of obtaining an optimal chromaticity for an ophthalmic tint, (1.38). A set of ophthalmic trial filters, included in the patent, has been designed in collaboration with Cerium Optical Group. The filters have one of seven hues, and there are five filters of each hue, with progressively increasing deposition of dye, each with a smooth spectral transmission. Using trial filters from only two of the seven hues, it is easy to obtain a combination of filters that resembles any chromaticity selected in the Colorimeter. Any chromaticity in a large gamut can be closely approximated in a filter that has a smooth spectral transmission, and that therefore minimises metamerism. Once a combination of trial filters has been selected, it can be ratified by the patient under normal viewing conditions. The Colorimeter and Trial lenses constitute a system for precision ophthalmic tinting. The system is marketed under license by Cerium Visual Technologies, together with a full instruction manual (1.116). The combination of trial lenses chosen by the patient forms a prescription that can be used by optometrists to specify the required spectral transmission. The dyeing is undertaken by Cerium. It is guided by eye, using a method which ensures that a visual match provides identical spectral transmission. The transmission is checked by a spectrophotometer and associated software (1.114) which provides quality control and prints a hand-out for the optometrist and the patient, describing various characteristics of their lenses, including the propensity to interfere with colour perception.
The system has been evaluated with children who have reading difficulties, initially in an open trial (1.15) and, more recently, in a double-blind trial (1.37). In the latter trial the subjects were provided in turn with two pairs of spectacles, each pair having similar colour. One pair of spectacles matched the optimal Colorimeter setting, and the other pair matched a suboptimal setting at which perceptual distortions were apparent. Headache and eye-strain were reduced with the optimal pair, even though subjects were quite unable to give a preference, or to say which pair best resembled the Colorimeter setting.
This finding is counter-intuitive and suggests that the tints have an efficacy over and above that of any placebo. According to the theory of visual discomfort, the tints change the pattern of excitation in the visual cortex. The colour that minimises perceptual distortion is that which reduces the excitation in hyperexcitable areas. When this colour is provided by spectacles the neural network is then less at risk of an inappropriate spread of excitation.
C. Reading
The effects of colour on reading and eye-strain can also be evaluated using plastic sheets placed over a page of text (overlays), although the optimal colour differs from that for ophthalmic tints. A set of overlays that can be used to sample colour space evenly and efficiently has been designed (1.35), and is now marketed by the Institute of Optometry Marketing together with a manual describing their use (1.117). Overlays can improve visual search by a small margin but their greatest benefit would appear to be in reducing the symptoms and signs of visual discomfort from prolonged reading (1.26).
D. Photosensitive Epilepsy
The extensive empirical data on photosensitive epilepsy which forms the basis of the theory of visual discomfort has been reviewed in a book chapter (1.73) and it forms the substance of a report commissioned by the Department of Trade and Industry on Photosensitive Epilepsy Associated with Playing Computer Games (1.118). The response of photosensitive patients to new forms of video technology is being studied (1.14). The work has also been reviewed for the benefit of sufferers (1.115).
E. Other Work on Visual Perception
The evaluation of the Cambridge low contrast gratings in various clinical contexts has continued (1.21). In addition, the visual processes involved in rehabilitation have been reviewed (1.60).
F. Industrial Collaboration
The Comfort 41 lens is being marketed by Cambridge Optical. The Intuitive Colorimeter has been manufactured in collaboration with Cambridge Industrial Design and Wilj International, and is now marketed under license by Cerium Visual Technologies. The MRC tinting system which incorporates the Intuitive Colorimeter and tinted trial lenses was launched in the summer of 1993 and is now in use in more than 40 optometric practices in the UK. It is being marketed in Europe, the United States and Australia. The Intuitive Overlays are marketed by Institute of Optometry Marketing and have already been reprinted.
FUTURE PROPOSALS
A. A General Unified Theory
One of the difficulties with the assessment of visual discomfort is that the assessment is necessarily subjective. It is unsatisfactory to rely on reports of sensations for which there is no external validation, and no objective criterion. We propose a series of investigations designed to establish objective correlates of visual discomfort and to advance our understanding of the physiological basis of the discomfort.
A1. Visual Evoked Potentials:
The theory of visual discomfort proposes that perceptual distortions arise from a cortical hyperexcitability. This hypothesis can most obviously be tested by using visual evoked potentials. Evoked potentials have been shown to be abnormal in migraine, and have a particularly high amplitude in response to flicker at epileptogenic frequencies. Recent work has highlighted abnormalities in dyslexia, although these remain controversial. Kuba and colleagues (personal communication) have used visual stimuli that can be seen only by virtue of their movement, and report that the potentials evoked by these stimuli may demonstrate unusually large abnormalities in a variety of clinical contexts, including dyslexia. It is proposed to study the potentials evoked by flicker, by pattern reversal and by pattern movement in patients who find tinted glasses beneficial. The potentials obtained when the stimulus has the preferred chromaticity will be compared with those obtained when the chromaticity is complementary and when the chromaticity is similar to those of conventional illuminants. These studies will require the purchase of a system for evoked potential monitoring. They offer the possibility of an objective correlate of physiological functions that at present can be measured only with subjective methods.
A2. Functional Magnetic Resonance Imaging:
In collaboration with colleagues at Cambridge University Department of Medicinal Chemistry, it is proposed to use fMRI to study the cerebral response to a variety of visual stimuli. We propose to measure regional cerebral blood flow in visual cortex when observers view epileptogenic patterns. If the patterns are presented in one lateral visual field, and the other field contains an innocuous pattern, the activation in the two hemispheres can be directly compared. The theory of visual discomfort predicts that migraineurs who experience consistently lateralised visual aura may show relatively increased blood flow when the stimulus is presented in the same field as that in which the aura occurs. If the metabolic response to epileptogenic stimuli is large, it would be instructive to study individuals without migraine who suffer pattern glare.
A3. Other Objective Correlates:
Three further avenues will also be explored in the hope of obtaining objective correlates of visual discomfort.
A3.1 Visual masking:
The visual stimuli that give rise to discomfort are effective at masking low-contrast targets (1.99). It is hoped to develop a clinical test of masking that will be useful in the assessment of patients who complain of discomfort.
A3.2 Pupil size:
Under photopic levels of luminance, pupil area has been shown to be determined largely by the scoptopic luminance. It is proposed to measure the effects of tinted lenses on pupil size to see whether this might provide one explanation of their benefit in reducing photophobia and increasing visual clarity.
A3.3 EMG:
EMG power recorded from temporalis muscle has been shown to correlate highly with reports of glare from light sources close to the visual axis. It is proposed to record EMG activity when patients observe stressful patterns and to do so when the patterns are illuminated by light of a chromaticity that reportedly reduces the stress. This study will require the system for electrophysiological monitoring mentioned in A1.
A4. Lighting:
The theory of visual discomfort has been applied to lighting design, but to date, the emphasis has been on the effects of the rapid temporal modulation of light (e.g. Wilkins, Nimmo-Smith, Slater & Bedocs, 1989). It is proposed to continue the work on flicker, but also to extend the theory to encompass the effects of the spectral power distribution.
A4.1 Interaction between flicker from displays and that from lighting:
The refresh frequency of computer displays has been increasing, bringing it closer to the frequency at which office lighting has its predominant modulation. It is possible that there exists an interaction between the modulation from both sources. Visual adaptation to the "beat" between two high-frequency sources can be demonstrated psychophysically providing evidence that the "beat" may be a potential source of discomfort. It is proposed to measure movements of the eyes across computer displays under lighting which does and does not exhibit the typical 100Hz modulation. This modulation can be removed using high-frequency fluorescent ballasts, and may reduce eye fatigue in offices where computer displays are in frequent use.
A4.2 Individual differences in colour constancy:
Preliminary data have indicated that, when looking through a coloured lens, some individuals are unable to tell that it is coloured, even when the lens is very strongly saturated. It is already evident that there are large individual differences in this respect. These findings suggest that some individuals are far more affected by changes in illuminant chromaticity than others. We hope to explore these individual differences to determine whether they relate to discomfort and to the effects of the chromaticity on spatial vision.
A4.3 Ultraviolet radiation:
Recent informal observations have suggested that, under double-blind conditions, observers can detect the difference between filters that block UVA below 400 nm and those that block below 380 nm. The discrimination is made not on the basis of colour appearance but of visual discomfort. If confirmed, this observation may be of relevance to the use of UV-blocking dyes in ophthalmic lenses and to the increasing prevalence of lamps that emit ultraviolet light.
B. Precision Ophthalmic Tinting
The MRC system for Precision Ophthalmic Tinting was launched commercially in the spring of 1994. There are preliminary indications that precision tinting may offer benefits in reducing the photophobia associated with a range of different disorders, including not only dyslexia but migraine, head injury, and photosensitive epilepsy. It is proposed to set up clinical trials, in collaboration with colleagues. The majority of trials will be open and, if the results appear promising, will be followed by double-blind trials.
B1. Double-Blind Trial of Precision Ophthalmic Tinting in Migraine:
The indications for precision ophthalmic tinting as a treatment for photophobia have already been reviewed (1.91). A double-blind trial in migraine has been designed and is awaiting funding.
B2. Precision Ophthalmic Tinting in Patients with Head Injury:
Many patients with closed head injury report photophobia, particularly immediately following the trauma. I have began an open trial to assess the clinical impact of precision tinting in these patients. If the initial results continue to be successful, I shall organise a clinic in collaboration with colleagues in optometry and in rehabilitation.
B3. Open Trial of Precision Ophthalmic Tinting in Photosensitive Epilepsy:
Thus far, 12 patients with photosensitive epilepsy have been offered precision tints. Six have benefited. It is proposed to extend this trial and to explore ways in which colorimetry, currently available only in optometric practice, can be made available in the clinical neurophysiology laboratory for the assessment of patients with epilepsy. I hope to establish a clinic, with Dr David Fish at the National Hospital, at which a colleague in optometry, Anita Lightstone, will assist with the assessment of patients.
C. Reading
The theory of visual discomfort has been applied to reading. The effects of high frequency modulation of light on the control of eye movements have been studied, but the theory predicts that repetitive spatial modulation may also have deleterious effects. It is proposed to explore the effects of such spatial modulation on ocular motor control. It is also proposed to investigate the visual factors that contribute to stress when children learn to read.
C1. Eye Movement:
Eye movements are often studied in response to isolated targets rather than the spatially ambiguous stimulus provided by text. It is now known that rapid eye movements made to one of two neighbouring objects in the periphery of vision are inevitably directed to a intermediate position between the objects. The implications of this so-called "global effect" for movements of the eyes across text are obvious. It is proposed to investigate movements of the eyes across text that has been set so as to enhance or reduce the information available at low spatial frequencies.
C2. Contrast of Letters:
Paper used in the production of children's reading books incorporates fluorescent dyes that absorb ultraviolet light and re-emit in the visible range, giving the paper a "whiter than white" appearance. This potentially increases the contrast of print, already high because of laser typesetting. I hope to measure the fluorescent response of typical dyes and any enhancement of contrast they produce under typical conditions of illumination, and thereby to assess the likely impact on contrast and, by implication, on children's reading.
C3. Group Tests for Overlays:
One hundred and fifty five normal 7-year old school children have been examined individually with coloured sheets of plastic placed on a page of text (overlays). Forty-five percent reported improved perception with a colour, and most have subsequently used the chosen overlay without prompting for six months. Now that the MRC overlays are commercially available, it is important to ensure they reach the children who can benefit from them. It is also important to minimise the demands on teachers' time. In order to reduce the amount of individual assessment that is necessary, we are developing group testing methods. Preliminary data suggest that the majority of children who will not benefit from overlays can be detected by group testing methods. It is proposed to refine the present methods and improve their sensitivity and selectivity.
C4. Referral Structure:
It seems likely that the overlays provide a way of selecting those children who may benefit from coloured glasses, but this has yet to be demonstrated formally. There is much work to be done refining the techniques for assessment.
D. Photosensitive Epilepsy
It is proposed to investigate the physiological basis of scotosensitive epilepsy, contrasting it with photosensitive epilepsy. Trials of precision tinting are proposed, together with a study of the effects of fluorescent lighting on photosensitive epilepsy.
D1. Scotosensitive vs. Photosensitive Epilepsy:
Certain patients with idiopathic epilepsy exhibit occipital spikes that are present only when the patient is unable to see patterned vision (so-called "fixation-off sensitivity"). Curiously, these patients are not sensitive to flicker. The division of magno- and parvo-cellular visual pathways may offer one possible explanation. In collaboration with colleagues at St Thomas' Hospital, the physiological mechanisms of fixation-off sensitivity will be explored and contrasted with those of photosensitive epilepsy.
D2. Open Trial of Precision Tinting. See B3 above
D3. High-Frequency Flicker and Epilepsy:
There has been little study of the effects of high-frequency flicker on epilepsy. It is proposed to record the EEG of patients who are undergoing routine clinical telemetry in the course of treatment for epilepsy, comparing the incidence of epileptiform EEG activity when the recording suite is illuminated by conventional fluorescent light, and when, under double-blind conditions, the lighting is controlled by high-frequency circuitry that removes the 100Hz flicker.
MOVEMENT CONTROL (Goodrich, Nimmo-Smith, Sellen, Tresilian, Wing)
Introduction
Five themes touching on the work of several APU scientists are covered under the heading of movement control. The first covers variability of the two elemental characteristics of movement, force and time. The next three themes cover the related skills of reaching, catching and grasping with position coming to the fore as a critical task element added to force and time. Standing balance is treated in the fifth section.
A. Elemental Variability
Movement results from muscle activity producing force of a given magnitude that acts on the mass of the limb, possibly with the mass of an object to be moved, for a certain duration. With a fixed starting position, fluctuations from one occasion to another in either force or duration will result in a range of end positions. This section is concerned with theoretical accounts of variability in these elemental aspects of movement.
A1. Force:
It is surprising, given the elemental status of force in movement, that there has been relatively little research into force control, and this is particularly true with regard to variability. APU research focuses on isometric force control because this simplifies measurement and also is likely to reduce the non-linearities in tension development by muscles undergoing length changes.
A1.1 Parallel Force Unit Model (PFUM) (Wing; Collaborator: Ulrich): Wing has completed a series of studies to examine the form of brief (300 ms) force impulses in terms of the mean and variability of the force-time function. A motivating factor for the work has been a mathematical model, PFUM, in which it is assumed that observed force arises from summing a large number of independent force units with variable times of activation relative to a single central command (1.27). The resulting rise and fall of force depends on both the form of the force-time function of the individual units and the distribution of their onset times. PFUM predicts scaleability of form; the basic shape of the force-time function should remain the same whatever the peak force. The model also predicts that the force-time variability function may show a local minimum at maximum force. These predictions have been supported in a series of experiments (1.57, 1.67) with analyses that included the novel use of centralised moments as force-time waveform shape descriptors.
A2. Timing: The timing of movement may be studied by asking subjects to make periodic responses to define a time series of inter-response intervals. A fundamental premise is that variability in the intervals reflects the structure of underlying timing mechanisms.
A2.1 Individual timing of series of movements (Nimmo-Smith, Wann, Wing; Collaborators: Vorberg, Bisiacchi):
In a theoretical review, Wing (1.68, 1.69) set out a number of linear models as extensions to his original two-stage model (in which it was assumed that inter-response intervals reflect two independent sets of underlying delays, one arising in a central timekeeper or clock, the other in motor implementation processes). One extension posited hierarchically organised timekeeping, with nested clocks responsible for timing groups of responses. It was shown on the basis of existing data that this extended model holds for the production of musical rhythms but is rejected in the case of another keyboard skill, typing.
This pattern of results suggests that a distinction should be made between musical performance skills, where response timing is an explicit goal, and skills in which timing is an emergent feature rather than a requirement of the task. This limitation on the applicability of hierarchical timing also applies to handwriting. Contrary to earlier suggestions in the literature, Wann and Nimmo-Smith (1.31) found that the relative durations of pen strokes change with writing size, which violates hierarchical timing. In a further study, however, Wann and Nimmo-Smith (1.32) found that writing on a smooth surface, which tends to make the pen skid faster over the surface, results in subjects using greater pen pressure. This has the effect of increasing friction, so keeping pen velocity -- and hence timing of strokes -- relatively constant. A number of other issues relating to handwriting variability were reviewed by Wing (1.2, 1.28, 1.29, 1.75, 1.77).
If musical rhythm production involves hierarchical organisation of a number of clocks, this might be expected to make greater demands on cognitive resources than regular, equal-interval tapping. This suggests that a given interval might be produced with greater variability in a rhythmic context than when that same interval occurs in an isochronous sequence, a prediction supported by a study by Wing and Bisiacchi. Moreover, the implication that cognitive load is greater in producing a rhythm was reinforced by the finding that the effect of context (rhythmic vs. isochronic) interacted with the performance of secondary tasks, such as backward counting.
A2.2 Ensemble timing (Wing, Allison, Woodburn):
An important concept in the hierarchical timing model is that information about overall rate is passed down through the hierarchically nested clocks. This constitutes a 'classic' information processing model in the sense that unobservable stages are assumed whose existence can only be inferred indirectly (from the statistical dependence structure of the data). There are however situations involving human timing where the timekeeping elements are directly observable, namely, ensemble timing.
An example of ensemble timing is rowing: for individual oarsmen in an eight, the task may be defined as producing oar-strokes at regular intervals in phase with the other crew members. The relevance of rowing to the hierarchical timing model stems from the fact that the stern-most oarsman ('Stroke') sets the rate for the others. Because the line of vision in a racing shell is restricted, it is likely that timing information propagates along the shell from one oarsman to the next. Indeed a preliminary study (1.93) showed a between-oarsman correlation structure indicating that the information passes down each side of the shell separately, with Stroke sitting at the apex of the branches of a hierarchy. A second study (1.58) also showed between-rower correlations but there were no differences according to side of the shell. One possible reason for the discrepancy is that the crew in the second study were more experienced. They may have based their coordination on multiple cues available to all rowers simultaneously which would have removed any tendency to a side-specific statistical structure to the interstroke intervals.
B. Reaching
Consider reaching for an object; movements of the fingers open the hand at the same time as shoulder and elbow movements carry the arm forward. Although there is no mechanical linkage between the fingers, shoulder and elbow (so that independent movements of hand and arm can be produced if desired), in reaching there is usually a tight temporal coupling between them. In asking what is the basis of such coordination, it should be noted that movement variability and, in particular trial-to-trial covariation, can provide important clues. This was a central thesis of Wing's tutorial review of 25 years of motor control research in Attention and Performance (1.43).
B1. Hand Positioning after CVA (Wing, Lough; Collaborators: Fraser, Turton, Jenner):
Moving the hand between two points in space requires various contrasting patterns of coordinated shoulder and elbow movements. For example, moving the arm across the front of the body involves simultaneous shoulder and elbow flexion, whereas a forwards movement starting at the body midline requires extension of the elbow with flexion of the shoulder. Clinical observation of arm movement after a CVA has led to the suggestion that flexor or extensor synergies predominate (so that midline movements are difficult to perform). In a longitudinal study of stroke patients, Wing and Lough (1.47) sought a quantitative description of this limitation on coordination. Motion of the elbow joint was examined in stroke patients when they made movements in a horizontal plane with arm supported against gravity. Targets were positioned so that flexion, extension or absence of motion at the shoulder was required. There was no reliable influence of shoulder motion on elbow movement, nor did this picture change with time following stroke onset. Thus there was no support for the fixed synergy view, and it may be that the clinical impression of stroke patients reflects salient changes in resting posture in response to gravity rather than alterations in coordination in active movement.
B2. Hand Shaping in Prehensile Reaching (Wing, Haggard; Collaborator: Jenner):
Picking something up requires an approach by the arm coordinated with hand opening to encompass the target object. Wing (1.74) reviewed a case study which showed that a user of an artificial hand employed a pattern of coordinating hand opening with arm movement that was the same in both the artificial and the natural hand. This similarity was observed despite quite different musculature being used to control the artificial hand. It suggests that action is specified in terms of motions in external space rather than being based on a joint/muscle representation of action.
With arm motion impaired in the artificial hand, it is interesting to note that the artificial hand opened relatively wider in approaching an object of given diameter. This suggests there was strategic compensation by the hand for limitations on arm movement. This was also reported in another single case study, this time of a cerebellar patient (1.9). Capitalising on the asymmetry of the patient's symptoms, Haggard and Wing showed that ataxic arm movement is associated with wider hand opening. An important double dissociation between the components was obtained when the patient was asked to reach in the dark. This improved position control of the ataxic arm but resulted in less accurate movement of the other arm. There were complementary changes in maximum hand aperture and this emphasised the strategic nature of coordination. A further example of the dependence of hand shaping on the transport component was found in a study in which normal subjects were asked to pick up an object after traversing an intermediate spatially specified "via" point (1.50, 1.59, 1.100). Compared to movements in which the subject was free to move the hand along a straight line path, shaping was significantly delayed.
The above experiments were based on naturalistic manipulation of hand/arm coordination. An alternative approach is to interfere with one component and assess effects on the other. An example of this approach involved innovative use of a linear actuator. The actuator was used to slow down the hand transport component of reaching by the arm to see whether and how the hand would adjust its shaping (1.11, 1.100). It was found that perturbation of transport typically elicited (after a delay comparable to a normal RT) a slowing or a reduction in hand aperture. A formal model with spatially-based coupling between hand and arm movement has been developed to account for these results (1.51, 1.100).
The measurement of movement in reaching can be a time consuming process when it is based on hand-digitised, frame-by-frame video analysis. Recent research at the APU has been considerably facilitated by a computer-based tracking system (Watsmart) although certain problems with the method had first to be surmounted (1.10)
B3. Bimanual Positioning Control for Computer Input (Sellen; Collaborator: Buxton):
The hypothesis underlying Sellen's work is that the sensorimotor configuration of human-computer interaction influences cognitive aspects of human performance (and vice versa). The broad goal is to understand the relationship between the sensorimotor and the cognitive aspects of interaction in order to develop innovative computer input techniques and devices. Sellen has examined compound tasks involving two-handed input (1.83). The starting point was Guiard's (1987) observation that in most everyday bimanual tasks, the actions of each hand tend to be different but interdependent. Three different bimanual techniques for accomplishing a computer task were compared: in one, the two hands controlled independent subtasks, and in two others the action of the right hand depended on that of the left hand. Results showed that one of the two dependent bimanual techniques resulted in significantly faster performance and lower cognitive load than either of the other techniques. The independent bimanual technique, however, was no faster and resulted in higher cognitive load than a conventional one-handed technique.
C. Interception
Many actions such as catching a ball require very accurate spatio-temporal coordination of the hand with a moving target object. In this section, we focus on what visual information is used to control hand motion in space and time. First, the stationary observer is treated; the ball's trajectory requires positioning of the arm only, and movements typically last a few hundred milliseconds. Second, the situation is considered in which there is motion of the whole body and this must be controlled relative to a moving target over periods of several seconds.
C1. Catching (Tresilian):
A new theory has been developed to account both for the timing skills shown in the execution of interceptive actions and for the timed responses and relative time-to-collision judgements obtained in laboratory tasks (1.25, 1.55, Tresilian, 1991). This new theory raises a host of new empirical questions. An attempt to use existing published data to address some of these has been attempted (1.22) though it was concluded that existing empirical studies are either inadequate for this purpose or have failed to control important factors.
Drawing on the early work of Lee (1976) as well as upon logical arguments, the theory proposes that interceptive actions are timed using perceptual information about time-to-collision (TTC). Geometric analysis has revealed that numerous sources of such TTC information are available to an observer, some of which are appropriate for object motion and others for observer motion (1.24). Thus, in distinction to Lee's work, the theory stresses multiple sources of TTC information, context- sensitive processes for selecting and combining the different information sources and the role of multiple sensory systems. Experimental work has confirmed some predictions based on these ideas (1.23).
In the new theory an important distinction is proposed between extrapolation and relative TTC judgement tasks (in which cognitive operations on the perceptual input are implicated) and fast interceptive actions (implicating a direct low-level coupling of perception and action which bypasses cognitive operations). A strategy for modelling the latter kind of process has been proposed and a detailed model for the visual regulation of timing has been developed using the equilibrium point hypothesis for motor control. The model has been used to simulate the effects of the following factors on the temporal precision and movement kinematics of interceptive actions: (i) the nature of the control scheme (continuous, intermittent or preprogrammed); (ii) neural transmission delays; (iii) low-pass filtering characteristics of the human visual system's response to image motion; (iv) perceptual threshold characteristics; (v) limb dynamics; and (vi) the sluggishness of muscle response to neural input signals due to the dynamics of the excitation-contraction coupling process.
C2. Servo Control Schemes for Braking and Intercepting (Tresilian):
The evidence from fast interceptive actions suggests that they are controlled by predictive information in the spatial domain as well as in the temporal domain discussed in the previous section. There are actions involving moving targets (or the observer moving relative to stationary targets) which evolve over periods of seconds rather than milliseconds, and it has been proposed that control of two such actions -- braking and running to catch a ball -- are controlled using a servo type strategy which does not involve predictive information (Lee, 1976; Chapman, 1968).
Both Lee's and Chapman's proposals involve the person attempting to maintain visually measurable quantities at fixed values throughout performance of the action; successful performance is guaranteed if the person is successful in keeping these quantities sufficiently close to their desired values. An obvious prediction, therefore, is that during braking or running to catch a ball, the relevant quantities will be close to their theoretically desired values throughout performance; this prediction has been confirmed both for braking and catching (e.g. Lee, Reddish & Rand, 1991; Dienes & McLeod, 1993). However, modelling of these tasks has revealed that this type of data is quite insufficient as a test of these servo control hypotheses because maintenance of the relevant quantities close to their theoretically desired values during performance is also a prediction of other braking and catching strategies which involve predictive information (1.56).
D. Grasping
If a subject is asked to move a grasped object, accelerating then decelerating the hand induces reactive forces on the object and these add to the gravitational load (i.e. object weight) exerted on the hand. The problem for the motor system is that the grip force used when holding an object at rest is not, in general, sufficient to withstand the additional reactive forces and so these may result in the object slipping. The psychological interest is that the necessary compensation can be anticipated as it is a function of perceivable object attributes (e.g. mass, friction) and the nature of the planned movement (e.g. acceleration level). The study of grasp during movement is a new area, significantly extending the 'classic' analysis (by the Umea group: Westling & Johansson 1984) of the loading phase of grasping an object. An important contribution to this research has been the development of a simple but reliable means of recording grip force and movement. This involved using an accelerometer attached to a force transducer so that acceleration and deceleration are measured orthogonal to the grip force produced by the precision grip acting on the force transducer.
D1. Anticipatory modulation of grip force (Wing, Tresilian; Collaborator: Flanagan):
Research at the APU on grip adjustments for rapid arm movements has shown that people anticipate self-induced loads and compensate for them by adjusting grip force up and down in parallel with the load forces induced by the arm kinematics (1.7, 1.8). The grip force adjustments are remarkably robust. Increases and decreases in grip force with modulation of acceleration persist throughout cyclic movement even though it might be thought simpler for the nervous system to raise the grip force to a new higher level (1.48). The adjustments in grip force are seen across a variety of grasps formed with different effector combinations and can also be observed when the acceleration arises from whole body and not just arm motion (1.6). This suggests that setting of grip force is muscle independent and is therefore likely to be represented at a relatively high level. However, the relevant processing module appears cognitively impenetrable in that people are apparently unable to exert conscious suppression of grip force adjustments associated with arm movement.
D2. Sensory Factors in Grasp (Wing, Allison; Collaborator: Flanagan):
The grip force required to hold an object depends on its surface roughness as well as on its weight. It is therefore interesting to ask whether weight judgements might be influenced by surface texture. In fact, a recent experiment (1.49) has shown that a smooth object is judged to be heavier than a rough object of equal weight. This suggests that efferent commands are used in weight judgement, and that the brain fails to differentiate efferent commands to the hand (responsible for grip force) and the arm (responsible for the lift force).
E. Standing Balance
The maintenance of upright posture involves an equilibrium of static and dynamic forces. The sum of external forces acting on the body may be represented as a vector through the body centre of mass (CM), which is located approximately in the middle of the body at waist height. If the vector projects outside the base of support provided by the feet, motion results and can lead to a fall. To counteract this possibility, advance compensatory action may be taken -- such as leaning to move the CM -- so that when the force vector direction changes, it continues to project within the base of support. The position of the CM is therefore of considerable importance to maintenance of balance.
E1. Weight Distribution (Wing, Barton):
When the only force acting on the CM is gravity, the distribution of the weight of the body between the feet is a good index of CM position. A special piece of apparatus, SwayWeigh, has been developed by Wing and Barton (1.123) to provide a portable, low-cost and reliable means of measuring weight distribution. The design of SwayWeigh allows it to be used both for biofeedback purposes and for clinical scientific measurement. A useful feature in respect of the latter application is that it can be switched to measure fluctuations in weight distribution (wobble) which reflect active adjustments to posture in maintaining balance. A further feature is that all measurements are electronically scaled to bodyweight which makes them easier to compare. SwayWeigh has now been taken up commercially.
E1.1 Compensating for changes in CM position (Wing, Barton; Collaborators: Turton, Howick):
Consider the situation in which one arm is held up and out to one side and kept in that position. Compared to arms by the side, the raised arm leads to a small displacement of the body CM towards the edge of the base of support. This reduces the safety margin for overbalancing. A large scale study was conducted at a British Association for the Advancement of Science science fair to address the interesting question of whether posture is adjusted for this change in static equilibrium conditions. The results showed that people do correct for the effect on left-right weight distribution of holding an arm out, probably by leaning away from the raised arm (1.44).
To explore this aspect of balance control further, another experiment has assessed weight distribution when subjects are asked to hold a 5 kg weight in one hand, arms by the side, which induces a shift in whole body CM position similar to raising one arm. Interestingly, the results were instruction-dependent. If asked to stand 'upright', subjects tolerated the resulting shift in weight distribution, probably using vestibular or visual cues to maintain posture. However, if asked to stand with 'even' weight distribution, subjects could do this, suggesting an ability to access cutaneous or proprioceptive information from the lower limbs.
E1.2 Balance in the elderly (Maylor, Wing):
An increased incidence of falls in the elderly is associated with decreased postural stability. Maylor and Wing have been studying the possible relation between standing stability and cognitive task performance in the elderly. Standing balance in middle-aged and elderly volunteers was measured using SwayWeigh in the anterior-posterior direction. As might be expected, the elderly were significantly less stable than the middle-aged. In addition, there was also evidence of greater between- and within-subject variability in the older group. However, there were qualitatively similar changes in stability across the two age groups as a result of (1) raising and lowering the arms, and (2) standing for a long time. In contrast to these additive effects of age on stability as a result of physical changes, performing certain additional cognitive tasks produced interactive effects. For example, differences in stability between the middle-aged and elderly increased significantly when subjects simultaneously performed the Brook's spatial memory task. In contrast, random digit generation did not produce such an interaction. This suggests that, although standing balance is normally a highly automatic function, in the elderly it draws on cognitive processing resources; when these must be shared, balance performance declines. From the pattern of tasks that interacted with balance, one possible candidate is the visuo-spatial sketch pad of working memory.
E2. Standing Balance after Stroke(Wing, Goodrich; Collaborators: Jenner, Burgess-Limerick, Clapp, Virji-Babul, van Vliet):
Impairment of sensorimotor function commonly results from a CVA and can have profound effects on standing balance. Our experience with SwayWeigh, combined with an interest in impairments in balance after stroke, led to a review of the literature which showed the potential benefits of weight-shift training using biofeedback from devices such as SwayWeigh (1.76). However, methods for assessing impaired stability after stroke are lacking. Research on standing balance usually involves perturbing balance by displacement of the support surface under the feet. A novel alternative technique which involves applying controlled horizontal forces at waist level has been developed (1.45). The method has the advantage of allowing the effects of application and removal of force to be examined separately, which is not the case with support surface displacements. It is also particularly appropriate to the study of balance after stroke because it affords the possibility of stabilising the patient after a push if the hip position indicates instability and possibility of falling.
An initial study (1.46) of stroke patients using sideways forces at waist level has shown that they are less stable, particularly when the induced sway is towards the more affected side. A comparison of force application and force removal showed that the latter caused more problems. The individual patients were also documented in terms of their performance of various functional balance tasks. To score these, a procedure was developed based on averaged ratings obtained from groups of judges. A number of issues concerning ratings of motor impairment are covered in a collaborative study by Goodrich (1.13).
Over trials, the stroke patients in the previous study tended to improve in their ability to cope with destabilising forces, which suggests therapeutic effects of having balance gently probed. In this respect it is relevant to note van Vliet and Wing's (1.30) suggestion that, if robotic devices could be engineered to assist voluntary movement, they might serve a rehabilitative function in allowing physically handicapped users to explore altered relations between action and sensory consequences.
FUTURE PROPOSALS
Introduction
The proposals for future work on movement control develop the five themes covered in the last 5 years. A new area of proposed work, on proximal and distal anticipatory postural adjustments, will explore commonalities between the stabilisation functions of grasp and standing balance.
A. Elemental Variability
Variability in the timing of muscle activity (which affects duration and relative phase) or in the level of activity (which affects force) is an important constraint on the accuracy of ballistic movement. There is a fundamental division in the study of movement timing, between central and peripheral processes acting as separate variance sources. Wing will extend this distinction to the production of brief force impulses, by adapting an experimental paradigm used in timing studies in which subjects make repeated simultaneous responses with two bilaterally symmetric effectors. When timing is the focus of interest, common variation in the (within-hand) intervals between left-left and right-right responses is taken to represent variability in a central process that sets up a common command to the two effectors (1.69). Variation in the (between-hand) difference between response times (left-right asynchrony) is taken to reflect variability in separate peripheral control mechanisms for the two effectors. By making force the focus, with a single target peak force for left and right hand responses, common variation in observed force may be taken to represent variability in a central process that sets up a force command that is common to the two effectors. Variation in the difference in force between response pairs is taken to reflect the sum of variability in two separate sets of low-level processes associated with response execution.
A1. Force
A1.1 Parallel Force Unit Model (PFUM) (Wing; Collaborator: Ulrich):
Work will continue with PFUM, which comprises a specific type of peripheral variation in force. In PFUM the underlying force elements are assumed to act unidirectionally; a force impulse results from the summation over a large number of elements, whose duration may vary, but whose contribution is all positive. However, consider an observation of zero force. This may either indicate there is no muscle activity or that there is activity in opposing muscle groups, each group contributing equal torques of opposite sign with a net result of zero force. The utilisation of opposing muscles in the generation of brief force pulses will be investigated by asking subjects to vary the degree of agonist-antagonist co-contraction. Pilot work has shown this may be achieved by asking the subject to stiffen the joint, using surface electromyography (EMG) to monitor muscle activity. The mean and variance functions of the force impulses will be examined in relation to the level of antagonist contribution. PFUM will be extended by the inclusion of negative force units whose recruitment may be coupled in varying degree to the positive units.
A1.2 Bilateral force impulses (Wing):
Subjects will be asked to hold two force transducers, one in each hand. In one paradigm, they will produce brief force impulses to match targets defined in terms of peak force and duration or impulse. In a second paradigm, subjects will maintain a steady force on which they superimpose step increases or decreases to match target force steps. In either paradigm the target will be kept constant over a series of trials. Serial statistics will be estimated including covariance in force measure between the hands (central variance) and variance of the difference in force produced by each hand (twice the peripheral variance). The two paradigms will be checked for consistency of the within-subject estimates of the two parameters in the model.
Tasks that load central and peripheral processes differentially will be employed to demonstrate dissociation of effects on the variance estimates. Loading of the central process will involve varying the target force between a number of remembered values during a series of trials. Peripheral component variance will be manipulated by varying the baseline force or by changing the effector pair. Dissociations between central and peripheral force processes will also be sought by asking subjects to carry out the bimanual force production task while performing concurrent secondary tasks that differentially affect central and peripheral factors.
A2. Timing
A2.1 Individual timing of bilateral responses (Wing):
The work in timing will parallel that described for bilateral force impulses. Vorberg and Hambuch's (1984) study of synchronous bimanual tapping will be replicated and extended by including manipulations designed to load (common) central or (lateralised) peripheral processes contributing to timing. The central component will be loaded by requiring rhythmic patterning of repetitive responding. Changes in the peripheral component will be induced by varying the movement distance. The limiting case is no movement in which "tapping" involves isometric force pulses. Investigation of this case will allow evaluation of the response, both in terms of the time intervals between responses and their force. Cross-correlations between force and timing components will be estimated while varying emphasis is placed on force and timing targets. The goal will be to evaluate the modularity of these two elemental components of motor control.
A key assumption in the proposed research on timing and force control is that bilateral simultaneous movements are triggered by a common command. This constitutes a hierarchical perspective on motor control, and may be contrasted with a distributed control view of simultaneous responding which holds that each effector is driven by its own separate control system, i.e., there is no single central process for force or timing (Yamanishi, Kawato & Suzuki, 1980; Wing, Church & Gentner, 1989). On the hypothesis that control may be hierarchical for simultaneous responding and distributed otherwise, experiments will be conducted to determine whether, with training, control can be moved between the two modes; this can be accomplished by starting with simultaneous responding and introducing small amounts of asynchrony, either by instruction or by varying effector combinations, for example between hand and foot. The proportion of common variation of within-hand time or force measures will be taken as an index of the degree of hierarchical control.
A2.2 Ensemble timing (Wing, Nimmo-Smith; Collaborators: Woodburn, Vorberg):
A concrete example of distributed control is ensemble performance in rowing. Individual crew members are clearly capable of making independent strokes; to achieve the observed similarity in force and timing of their movements, there must be transfer of information between rowers. Models for ensemble timing will be developed which should provide a useful base for understanding within-individual coordination between limbs. An important component of the work will be the development of analytic tools for characterising short- and long-term fluctuations in performance within and between individuals. The performance of rowing eights of various levels of expertise will be recorded with the goal of determining the balance between the use of local information (e.g. to maintain relative phase with the rower in front), global information (e.g. to maintain relative phase with motion of the shell) and prediction (e.g. using the verbal cue provided by the cox on the required stroke rate to generate a target interval on the basis of past experience). A further issue that will be investigated is how individuals resolve the competing demands of minimising interstroke interval variance or minimising synchronisation variance.
A2.3 MRI and rhythm production (Wing; Collaborator: Hall):
In single-effector timing, the central-peripheral distinction has been successfully applied in neuropsychological studies of timing deficits in patients with motor disorders. Thus, Ivry, Keele and Dienes (1988) have argued that circuitry underlying central timing processes involves the cerebellum, but not the basal ganglia or motor cortex. To test further the idea that timing is a cerebellar function, an fMRI study will be performed at the Addenbrooke's Hospital facility with normal volunteers, to determine whether there is altered patterning of activity in the cerebellum when repetitive (unilateral) movements must be accurately timed with equal intervals. The contrast of interest will be with similar movements produced in reaction to randomly timed external stimuli at the same overall rate. If this is successful, a further fMRI experiment will seek to contrast simple repetitive timing of a single interval with the production of a rhythmic pattern. Luria's (1966) work on rhythm deficits after stroke suggests that the patterning of time in complex rhythmic tasks is a function of the left frontal cortex. If so, rhythm production in normals should reveal heightened localised cortical activity in that area in addition to elevated activity of the cerebellum as seen in equal interval tapping. Again it will be important to control the overall rates of responding for comparability of baseline metabolic activity levels.
B. Reaching
B3. Two-handed Cooperation (Sellen, Wing, Bishop; Collaborator: Buxton):
Various theoretical accounts of one-handed positioning movements have been proposed (e.g. Meyer et al., 1990) but little is known about how the models apply to two-handed action. Interference is known to occur between simultaneous hand movements with independence as the task goal (e.g. Kelso, Putnam & Goodman, 1984) but the interesting issue of how the two hands are controlled in cooperative action has not been studied systematically. Sellen, Wing and Bishop propose to document such movements, with particular interest in laterality effects. Building on the anecdotal observations of Guiard (1987), two-handed action in selected naturalistic tasks (e.g. picking up and sorting scattered playing cards by suit, opening and closing containers, assembling Lego building blocks) will be videotaped. A detailed, descriptive analysis of the action of each hand in relation to the other will focus on sharing of stabilisation and manipulation aspects of movement. To gain quantitative insight in the mutual dependence of the two hands, Watsmart will be used to record two-handed picking up of various types of objects between the flat palms. Of interest will be the relative guidance of left and right hands (defining a notional grasp shape in approach) under various levels of difficulty and visual constraint.
Practitioners in the human-computer interaction community are beginning to take a keen interest in two-handed interaction techniques, and Sellen will also investigate the kinds of interactions that might be redesigned for two hands. In a study which examined non-preferred hand performance in a one-handed aiming task, Kabbash, Mackenzie and Buxton (1993) found that performance was faster and more accurate with a mouse compared to a trackball. However, in a pilot study that involved asymmetric, dependent action of both hands, Sellen has found that the speed and accuracy of the non-preferred hand is unaffected by device (trackball or mouse) even though it was performing the same aiming movements as in the one-handed case. Sellen plans to pursue this finding in a more complete experiment, comparing performance of the left hand in one- and two-handed tasks. If factors such as device make a difference in the one-handed task, but not in the two-handed task, this will suggest that the subject's conceptualisation of the task can have significant effects on the articulation of the action.
C. Interception (Tresilian)
The derivation of alternative models, plus simulation studies designed to determine how best to distinguish the models empirically, will continue to be an important part of Tresilian's study of interception tasks. Various models for braking and catching have already been put forward and it has been shown how these alternatives can all account for existing empirical data. Simulation studies of these models have revealed how various control schemes behave and have also revealed that some control schemes are more desirable than others due to their insensitivity to noise and variation in dynamic parameters. It is proposed to follow up the theoretical papers with empirical work to distinguish between the various models.
D. Grasping
D1. Anticipatory Modulation of Grip Force (Wing, Tresilian; Collaborator: Flanagan):
Previous APU research has demonstrated a robust linkage between increases in grip force and increases in inertial load force during rapid movements of an object held in the hand with a precision grip. Because of their early onset, the grip force adjustments may be identified as anticipatory. What is the basis for this anticipatory organisation? Familiarity with the object's mass and frictional characteristics as a result of slips (which can be identified from the acceleration trace) on previous trials is probably not the answer, since anticipatory grip force modulation is typically seen on the very first trial. Another possibility is that subjects sense the mass in the initial hold phase of the trial, from which they can then compute an expected load force given any chosen acceleration. Given that they are also able to assess surface friction, they may then modulate grip force appropriately.
In a study of the basis of these anticipatory processes, subjects will be asked to move 'virtual' objects endowed with contrasting dynamic properties. The manipulandum will be mounted on a 2-axis force-servoed linear motor system newly-developed at the APU. The system is like a large x-y pen plotter, but it allows any desired force field (e.g. a spring load in which restoring force increases with distance from an equilibrium position) to be defined by computer. The issue is then how grip force modulates over successive trials as subjects learn the dynamics associated with any particular virtual object. Changes in anticipatory grip-force modulation may also be accompanied by changes in the responsiveness of reflexes that can provide recovery of stability if a slip should unexpectedly occur. This possibility will be evaluated by introducing occasional unexpected momentary obstructions to movement while recording surface EMG from the muscles contributing to grip force.
D2. Sensory Factors in Grasp (Wing, Carlyon; Collaborator: Lederman):
The skin on the pads of the fingers and palm of the hand is endowed with numerous sensory receptors which vary both in their rate of adaptation and in the size of their receptive fields (Johannson & Westling, 1990). The properties of these receptors have been charted out with traditional psychophysical procedures adopted from audition, with appropriate change of frequency scale. Their activity has also been demonstrated during functional movement and they have been implicated in the reflex control of grip force to prevent an object slipping from grasp. However, little is known about sensation of a grasped object during voluntary movement. By analogy with saccadic suppression, cutaneous sensory thresholds might be expected to rise during movement. Yet this is just the time when slip information might be needed for reflex support of anticipatory grip adjustments. Wing and Carlyon, in collaboration with Lederman, propose to evaluate sensitivity to vibration of a grasped object during voluntary movement. Testing will include both monitoring of the activity of muscles subserving grasp, and forced-choice verbal report in order to test for dissociation between the effects of movement on grip force and the ability to detect which of two sequential observation periods contained the vibratory signal. If such a dissociation were obtained, it would extend Milner and Goodale (1993) important distinction between vision processing for conscious report versus action into the haptic domain.
D3. Proximal and Distal Anticipatory Postural Adjustments:
Psychologists studying control of movement often distinguish between fine motor skills such as writing, which involve distal musculature (usually of the hand), and gross motor skills such as kicking a ball, which involve musculature of the whole body. However, certain similarities in task definition suggest that this distinction is somewhat contrived. Thus, the role of the digits in modulating grip force to stabilise an object that is subject to external or self-induced loads has functional parallels with action of the legs and trunk that serves to maintain stable upright posture in the face of forces imposed by the environment or by one's own movements.
D3.1 Normal (Wing, Nimmo-Smith; Collaborators: Flanagan, Richardson):
While holding an instrumented load cell in the hand and standing with the feet on separate force plates, subjects will be asked to move the arm rapidly so that the temporal correspondence between grip force adjustment and whole body postural adjustments (as measured by the ground reaction forces) can be examined on a trial by trial basis. Watsmart will be used to track segmental motions. Surface EMG will further document the extent to which there is a correlation between adjustments which would imply common control principles, or even a common controller. Other conditions to explore this possibility will include: (i) jumping on the spot while holding the load cell in fixed position relative to the body, producing contrasting loads on take-off and landing; (ii) using the load cell to push against a fixed support to disturb balance; and (iii) pressing a button to trigger a force that pushes on the load cell and destabilises grasp and standing balance. An important element of this work will be the development of biomechanical models for the forward dynamics, which show how forces at individual joints contribute to observed forces at the end-effector.
D3.2 Changes in cerebellar and Parkinson's disease (Wing; Collaborator: Sagar):
If there are correspondences in the control of stability of grasp and standing balance in normals, it is possible that there is a common underlying process. One likely candidate for such an anticipatory coordination mechanism is the cerebellum, incoordination of upper and lower limb (ataxia) being a 'classic' cerebellar sign. Cerebellar patients will be studied for evidence of breakdown in coordination in anticipatory postural adjustments and grip force adjustments for rapid voluntary arm movements. Comparisons will be made with Parkinson's patients, whose impairment of postural and grip force responses is expected to be restricted to unexpected external perturbations, while their anticipatory adjustments will appear relatively unaffected.
E. Standing Balance
E1. Weight Distribution
E1.2 Balance in the elderly (Maylor, Allison, Wing; Collaborators: Quinn, Simpson):
The explanation suggested for age-related increases in interference between maintaining postural stability and performing cognitive tasks is that cognitive processing resources may be taken by standing balance. Further work is necessary to understand the pattern of effects observed across the different indices of postural stability obtained from SwayWeigh. For example, some cognitive tasks appear to increase random variation (wobble), while others cause a slow drift in weight distribution. To explore the microstructure of such task effects on stability, a reaction time procedure has been developed and pilot experiments performed. This procedure allows the investigation of random and systematic effects of stimulus processing on sway during the response period. Cognitive task performance may also be analysed as a function of the sway recording. Three lines of research will be explored with this paradigm. First, interference effects will be sought in younger subjects exposed to standing under more difficult conditions, such as eyes closed. Second, cognitive task demands will be manipulated by changing the content (e.g. spatial vs. verbal materials) and varying the speed requirements. Third, the possibility of voluntarily allocating resources differentially across the two tasks will be examined, as this will have important implications for training programmes designed to reduce falls in the elderly.
E2. Standing Balance after Stroke
E2.1 Neurophysiological changes and physical therapy (Wing; Collaborators: Jenner, Kirker; supported by Stroke Association):
Hemiparesis often renders stroke patients impaired in their ability to transfer body weight to or from the affected leg. This results in a risk of falling while standing when exposed to forces that may be external in origin or arise from self-initiated actions (including the initiation of gait). The technique developed at the APU of introducing horizontal forces at the hips will be used to evaluate the neurophysiological organisation of balance responses after stroke. Of particular interest will be coordination of the hip abductors and adductors. Pilot studies with a stroke patient many years post-insult have shown reorganisation of these bilaterally arranged muscles such that the non-affected side adductor takes over the function of the affected side abductor. If confirmed in other patients, this will have implications for the retraining of balance. The research will use behavioural and surface EMG responses to document short-term (acquisition) and long-term (retention) effects of training on stability for both expected (self-triggered) and unexpected perturbations. Transfer of training to walking will be evaluated in terms of executing a single step forward, since this requires anticipatory postural adjustments under voluntary control to effect the transition from stable standing to dynamic balance.
E2.2. Assessment of the 'pusher' syndrome (Goodrich; Collaborator: Ashburn):
In the majority of cases, stroke patients compensate for their balance difficulties by shifting their weight onto the ipsilesional (non-paretic) limb. However, in a few cases, patients express fear when encouraged to stand with their weight evenly distributed or predominantly on the non-paretic limb and insist on 'pushing' onto their paretic limb, frequently with disastrous consequences. In the clinical domain, these people are informally defined as 'pushers' and recognised as poor rehabilitation prospects. An assessment procedure has been devised whereby these patients can be distinguished from patients with other balance problems. Work will continue on establishing the reliability and validity of this assessment procedure and factors underlying this syndrome will be explored.
E2.3 Procedural learning after brain damage (Robertson, Wing):
Stroke is a complex disorder and it is likely that patients' limitations in taking benefit from biofeedback in relearning balance skills will reflect not only the hemiparesis but also central decision making and sensory factors. Robertson and Wing therefore propose to study procedural learning of simple skills with the non-affected side. This will allow investigation of the effects of training schedules, frequency and immediacy of knowledge-of-results on learning. Retention and transfer will also be assessed. It is also proposed to evaluate factors, such as verbal distraction during acquisition, which are likely to impede learning and which are often poorly controlled in hospital rehabilitation settings.
PUBLICATIONS (Excluding work done prior to arrival at APU)
Authored Books
1.1. WILKINS, A.J. (in press). Visual Stress. Oxford: Oxford University Press.
Edited Books
1.2. Wann, J., WING, A.M. & Søvik, N. (Eds.). (1991). Development of Graphic Skills: Research, Perspectives and Educational Implications. London: Academic Press.
Refereed Articles
1.3. Chen, E.Y.H., WILKINS, A.J. & McKenna, P.J. (1994). Semantic memory is both impaired and anomalous in schizophrenia. Psychological Medicine, 24, 193-202.
1.4. CHRONICLE, E. & NIMMO-SMITH, I. (1992). Application of some statistical methods for comparing samples of hue-angle data. Color Research and Application, 17, 375-378.
1.5. CHRONICLE, E.P. & WILKINS, A.J. (1991). Colour and visual discomfort in migraineurs. The Lancet, 338, p. 890.
1.6. Flanagan, J.R & TRESILIAN, J. (in press). Grip-load force coupling: A general control strategy for transporting objects. Journal of Experimental Psychology: Human Perception and Performance.
1.7. Flanagan, J.R., TRESILIAN, J. & WING, A.M. (1993). Coupling of grip force and load force during arm movements with grasped objects. Neuroscience Letters, 152, 53-56.
1.8. Flanagan, J.R. & WING, A.M. (1993). Modulation of grip force with load force during point-to-point arm movements. Experimental Brain Research, 95, 131-143.
1.9. HAGGARD, P., Jenner, J. & WING, A. (1994). Coordination of aimed movements in a case of unilateral cerebellar damage. Neuropsychologia, 32, 827-846.
1.10. HAGGARD, P. & WING, A.M. (1990). Assessing and reporting the accuracy of position measurements made with optical tracking systems. Journal of Motor Behavior, 22, 315-321.
1.11. HAGGARD, P. & WING, A.M. (1991). Remote responses to perturbation in human prehension. Neuroscience Letters, 122, 103-108.
1.12. Hazell, J. & WILKINS, A.J. (1990). A contribution of fluorescent lighting to agoraphobia. Psychological Medicine, 20, 591-596.
1.13. Henderson, L., Kennard, C., Crawford, T., Day, S., Everett, B., GOODRICH, S., Jones, F. & Park, D.M. (1991). Scales for rating motor impairment in Parkinson's disease: Studies of reliability and convergent validity. Journal of Neurology, Neurosurgery and Psychiatry, 54, 18-24.
1.14. Kasteleijn-Nolst Trenité, D., Dekker, E., Spekreijse, G., Brekelmans, G., WILKINS, A.J. & van Emde Boas, W. (in press). The role of television, video games and computers in epileptic photosensitive patients: Preliminary results. Epilepsia (Abstract only).
1.15. MacLachlan, A., Yale, S. & WILKINS, A. (1993). Open trial of subjective precision tinting: A follow-up of 55 patients. Ophthalmic and Physiological Optics, 13, 175-178.
1.16. PATTERSON, R.D. (1990). Auditory warning sounds in the work environment. Philosophical Transactions of the Royal Society, B.327, 485-492.
1.17. PATTERSON, R.D. (1990). The tone height of multiharmonic sounds. Music Perception, 8, 203-214.
1.18. PATTERSON, R.D. (in press a). The sound of a sinusoid: Spectral models. Journal of the Acoustical Society of America.
1.19. PATTERSON, R.D. (in press b). The sound of a sinusoid: Time-interval models. Journal of the Acoustical Society of America.
1.20. Smith, A.P., Tyrrell, D.A.J., Barrow, G.I., Higgins, P.G., Bull, S., Trickett, S. & WILKINS, A.J. (1992). The common cold, pattern sensitivity and contrast sensitivity. Psychological Medicine, 22, 487-494.
1.21. Somazzi, L., Della Sala, S. & WILKINS, A.J. (1990). A simple test of contrast sensitivity in glaucoma. Italian Journal of Ophthalmology, 4, 209-213.
1.22. TRESILIAN, J.R. (1993). Four questions of time to contact: A critical examination of research on interceptive timing. Perception, 22, 653-680.
1.23. TRESILIAN, J.R. (1994). Approximate information sources and perceptual variables in interceptive timing. Journal of Experimental Psychology: Human Perception and Performance, 20, 154-173.
1.24. TRESILIAN, J.R. (in press). A note on "Gravity as a monocular cue for perception of distance and/or absolute size. Perception.
1.25. TRESILIAN, J.R. (in press). Perceptual and motor processes in interceptive timing. Human Movement Science.
1.26. Tyrrell, R., Holland, K., Dennis, D. & WILKINS, A.J. (in press). Coloured overlays, visual discomfort, visual search and classroom reading. Journal of Research in Reading.
1.27. Ulrich, R. & WING, A.M. (1991). A recruitment theory of force-time relations in the production of brief force pulses: The parallel force unit model. Psychological Review, 98, 268-294.
1.28. van Galen, G.P., Thomassen, A.J.W.M. & WING, A.M. (1991). Handwriting: A movement theme. Human Movement Science, 10, 163-164.
1.29. van Galen, G.P., Thomassen, A.J.W.M. & WING, A.M. (Eds.). (1991). Handwriting. Human Movement Science (Special Issue), 10. Amsterdam: Elsevier Science Publishers.
1.30. van Vliet, P. & WING, A.M. (1991). A new challenge: Robotics in the rehabilitation of the neurologically motor impaired. Physical Therapy (Special series: Movement Science, Part 2), 71, 39-47.
1.31. Wann, J.P. & NIMMO-SMITH, I. (1990). Evidence against the relative invariance of timing in handwriting. The Quarterly Journal of Experimental Psychology, 42A, 105-119.
1.32. Wann, J. & NIMMO-SMITH, I. (1991). The control of pen pressure in handwriting: A subtle point. Human Movement Science, 10, 223-246.
1.33. WILKINS, A.J. (1991). Visual display units versus visual computation. Behaviour and Information Technology, 10, 515-523.
1.34. WILKINS, A.J. (1993). Health and efficiency in lighting practice. Energy, 18, 123-129.
1.35. WILKINS, A.J. (1994). Overlays for classroom and optometric use. Ophthalmic and Physiological Optics, 14, 97-99.
1.36. WILKINS, A.J. & CLARK, C. (1990). Modulation of light from fluorescent lamps. Lighting Research and Technology, 22, 103-109.
1.37. WILKINS, A.J., Evans, B., Brown, J., Busby, A., Wingfield, A., JEANES, R. & Bald, J. (1994). Double-blind placebo-controlled trial of precision spectral filters in children who report perceptual distortion. Investigative Ophthalmology and Visual Science, 35, p. 1573.
1.38. WILKINS, A.J., MILROY, R., NIMMO-SMITH, I., Wright, A., Tyrrell, R., Holland, K., Martin, J., Bald, J., Yale, S., Miles, T. and Noakes, T. (1992). Preliminary observations concerning treatment of visual discomfort and associated perceptual distortion. Ophthalmic and Physiological Optics, 12, 257-262.
1.39. WILKINS, A.J. & Neary, C. (1991). Some visual, optometric and perceptual effects of coloured glasses. Ophthalmic and Physiological Optics, 11, 163-171.
1.40. WILKINS, A.J., NIMMO-SMITH, I. & Jansons, J. (1992). Colorimeter for the intuitive manipulation of hue and saturation and its role in the study of perceptual distortion. Ophthalmic and Physiological Optics, 12, 381-385.
1.41. WILKINS, A.J., Peck, A. & Jordan, B. (1991). Visual discomfort in the classroom. Child Language, Teaching and Therapy, 7, 326-340.
1.42. WILKINS, A.J. & Wilkinson, P. (1991). A tint to reduce eye-strain from fluorescent lighting? Preliminary observations. Ophthalmic and Physiological Optics, 11, 172-175.
1.43. WING, A.M. (1992). The uncertain motor system: Perspectives on the variability of movement. In D.E. Meyer & S. Kornblum (Eds.), Attention and Performance XIV: Synergies in Experimental Psychology, Artificial Intelligence, and Cognitive Neuroscience (pp. 708-744). Cambridge, MA: The MIT Press.
1.44. WING, A.M., BARTON, J., TURTON, A. & Howick, I. (1992). Regulation of lateral position of body centre of mass in standing balance. Physiotherapy Theory and Practice, 8, 131-135.
1.45. WING, A.M., CLAPP, S. & Burgess-Limerick, R. (in press). Standing stability in the frontal plane determined by lateral forces applied to the hip. Gait and Posture.
1.46. WING, A.M., GOODRICH, S., VIRJI-BABUL, N., Jenner, J.R. & CLAPP, S. (1993). Balance evaluation in hemiparetic stroke patients using lateral forces applied to the hip. Archives of Physical Medicine and Rehabilitation, 74, 292-299.
1.47. WING, A.M., Lough, S., TURTON, A., Fraser, C. & Jenner, J.R. (1990). Recovery of elbow function in voluntary positioning of the hand following hemiplegia due to stroke. Journal of Neurology, Neurosurgery and Psychiatry, 53, 126-134.
Submitted
1.48. Flanagan, J.R. & WING, A.M. Grip force in cyclic movments. (Manuscript submitted to Experimental Brain Research).
1.49. Flanagan, J.R., WING, A.M., Allison, S. & Spenceley, A. Effects of surface texture on weight perception when lifitng objects with a precision grip. (Manuscript submitted to Perception and Psychophysics).
1.50. HAGGARD, P. & WING, A.M. Coordination of hand aperture with the spatial path of hand transport. (Manuscript submitted to Acta Psychologica).
1.51. HAGGARD, P. & WING, A.M. Coordinated responses following mechanical perturbation of the arm during prehension. (Manuscript submitted to Experimental Brain Research).
1.52. MCKEOWN, J.D. & PATTERSON, R.D. The time course of auditory segregation: concurrent vowels that vary in duration. (Manuscript submitted to Journal of the Acoustical Society of America)
1.53. ROBINSON, K. & PATTERSON, R.D. (submitted a) Is pitch required in the extraction of timbre? (Manuscript submitted to Journal of the Acoustical Society of America)
1.54. ROBINSON, K. & PATTERSON, R.D. (submitted b) Is pitch required to extract the timbre of steady-state instruments? (Manuscript submitted to Music Perception)
1.55. TRESILIAN, J.R. (submitted a). Automatic and cognitive processes in time-to-contact estimation: Analysis of the disappearing target paradigm. (Manuscript submitted to Perception and Psychophysics).
1.56. TRESILIAN, J.R. (submitted b) Study of a servo control strategy for projectile interception. (Manscript submitted to Quarterly Journal of Experimental Psychology).
1.57. Ulrich, R. WING, A.M. & Rinkenauer, G. Scaling of brief force impulses. (Manuscript submitted to Journal of Experimental Psychology: Human Perception and Performance).
1.58. WING, A.M & Woodburn, C. Coordination between rowers in a racing eight. (Manuscript submitted to Journal of Sports Sciences).
Invited Chapters and Commentaries
1.59. HAGGARD, P. (1992). Multi-sensory control of coordinated movement. In J. Summers (Ed.), Approaches to the Study of Motor Control and Learning (pp. 195-231). Amsterdam: Elsevier Science Publishers B.V.
1.60. Kennard, C. & WILKINS, A.J. (1993). Special senses. In R.J. Greenwood, M.P. Barnes, T.M. McMillan & C.D. Ward (Eds.), Neurological Rehabilitation (pp. 259-267). Edinburgh: Churchill Livingston.
1.61. PATTERSON, R. (1990). Auditory warning sounds in the work environment. In D.E. Broadbent, A.D. Baddeley & J.T. Reason (Eds.), Human Factors in Hazardous Situations (pp.485-492). Oxford: Clarendon Press. (Phil. Trans. R. Soc. London, B327).
1.62. PATTERSON, R.D. (1990). Listening. Brain, 113, 1245-1246.
1.63. PATTERSON, R.D. & AKEROYD, M. A. (in press). Time-interval patterns and sound quality. In G. Manley, G. Klump, C. Koppl, H. Fastl, & H. Oeckinghaus (Eds.), In Advances in Hearing Research: Proceedings of the 10th International Symposium on Hearing, Singapore: World Scientific.
1.64. PATTERSON, R.D., ALLERHAND, M. & HOLDSWORTH, J.W. (1993). Auditory representations of speech sounds. In M. Cooke & S. Beet (Eds.), Visual Representations of Speech Signals (pp. 307-314). Chichester: John Wiley & Sons.
1.65. PATTERSON, R.D. & HOLDSWORTH, J. (in press). A functional model of neural activity patterns and auditory images. In W.A. Ainsworth (Ed.), Advances in Speech, Hearing and Language Processing, vol. 3. London: JAI Press.
1.66. PATTERSON, R.D., ROBINSON, K., HOLDSWORTH, J., MCKEOWN, D., ZHANG, C. & ALLERHAND, M. (1992). Complex sounds and auditory images. In Y. Cazals, K. Horner & L. Demaney (Eds.), Auditory Physiology and Perception (9th International Symposium on Hearing, 1992) (pp. 429-446), Oxford: Pergamon Press.
1.67. Ulrich, R. & WING, A.M. (1993). Variability in brief force pulses. In K.M. Newell & D.M. Corcos (Eds.), Variability and Motor Control. Champaign, Illinois: Human Kinetics.
1.68. Vorberg, D. & WING, A. (1994). Modelle fur Variabilitat und Abhangigkeit bei der zeitlichen Steuerung. In Enzyklopaedie der Psychologie Gottingen (pp. 223-320). Germany: Hogrefe Verlag fur Psychologie.
1.69. Vorberg, D. & WING, A. (in press). Modeling variability and dependence in timing. In H. Heuer & S.W. Keele (Eds.), Handbook of Perception and Action, Vol. 3: Motor Skills. London: Academic Press.
1.70. WILKINS, A.J. (1990). Stress and distress from fluorescent lighting. In S. Puglisi-Allegra & A. Oliverio (Eds.), Psychobiology of Stress (pp. 211-221). The Netherlands: Kluwer Academic Press.
1.71. WILKINS, A.J. (1991). Visual discomfort and reading. In J.F. Stein (Ed.), Vision and Visual Dyslexia (pp. 155-170). Basingstoke: Macmillan Press.
1.72. WILKINS, A.J. (1993). Reading and visual discomfort. In D.M. Willows, R.S. Kruk & E. Corcos (Eds.), Visual Processes in Reading and Reading Disabilities (pp. 345-356). Hillsdale, N.J.: Lawrence Erlbaum Associates.
1.73. WILKINS, A.J., Binnie, C.D., Darby, C.E. & Kasteleijn-Nolst Trenité, D. (1990). Inferences regarding the visual precipitation of seizures, eye strain, and headaches. In M. Avoli, P. Gloor, G. Kostopoulos & R. Naquet (Eds.), Generalised Epilepsy: Neurological Approaches (pp. 314-326). Boston: Birkhauser.
1.74. WING, A.M. (1990). Coordination in normal and prosthetic reaching. In S.T. Venkataraman & T. Iberall (Eds.), Dextrous Robot Hands (pp.55-65). New York: Springer-Verlag Inc.
1.75. WING, A.M. (1990). Étude de la Variabilité dans la Forme Spatiale de L'Écriture Cursive. In C. Sirat, J. Irigoin & E. Poulle (Eds.), L'Écriture: Le Cerveau, L'Ceil et la Main Bibliologia Vol. 10 (pp.127-137). Turnhout: Brepols.
1.76. WING, A.M., ALLISON, S. & Jenner, J.R. (1993). Retaining and retraining balance after stroke. In C.D. Ward (Ed.), Baillière's Clinical Neurology, Vol. 2: Rehabilitation of Motor Disorders (pp. 87-120). London: Baillière Tindall.
1.77. WING, A.M., WATTS, F. & Sharma, V. (1991). Developmental dynamics of handwriting: Appraising the relation between handwriting and personality. In J. Wann, A. Wing & N. Søvik (Eds.), Development of Graphic Skills: Research, Perspectives and Personal Implications (pp. 164-175). London: Academic Press.
Conference Proceedings
1.78. ALLERHAND, M., BUTTERFIELD, S., CUTLER, A. & PATTERSON, R. (1992). Assessing syllable strength via an auditory model. In Proceedings of the Institute of Acoustics, Vol. 14, part 6 (pp. 297-304).
1.79. ALLERHAND, M. & PATTERSON, R. (1992) Correlograms and auditory images. In Proceedings of the Institute of Acoustics, Vol. 14, part 6, (pp. 281-288).
1.80. Anderson, T. & PATTERSON, R.D. (in press). Speaker recognition with the auditory image model and self-organising feature maps: A comparison with traditional techniques. In Proceedings of ESCA workshop. Martingy, Switzerland.
1.81. HOLDSWORTH, J., Schwartz, J.L., Berthommier, F. & PATTERSON, R.D. (1992). A multi-representation model for auditory processing of sounds. In Y. Cazals, K. Horner & L. Demaney (Eds.), Auditory Physiology and Perception (Proceedings of the 9th International Symposium on Hearing, 1992) (pp. 447-453). Oxford: Pergamon Press.
1.82. Irino, T & PATTERSON, R.D. (in press). A theory of asymmetric intensity enhancement. in non-simultaneous masking. In Proceedings of the Third ICSLP, Yokohama, Japan.
1.83. Kabbash, P., Buxton, W. & SELLEN, A. (1994). Two-handed input in a compound task. Proceedings of SIGCHI T94 (pp. 417-423). Boston, MA.
1.84. PATTERSON, R.D., Anderson, T. & ALLERHAND, M. (in press). The auditory image model as a preprocessor for spoken language. In Proceedings of the Third ICSLP, Yokohama, Japan.
1.85. PATTERSON, R.D., HOLDSWORTH, J. & ALLERHAND, M. (1992). Auditory models as preprocessors for speech recognition. In M.E.H.. Schouten (Ed.), The Auditory Processing of Speech: From the Auditory Periphery to Words (pp. 67-83). Berlin: Mouton de Gruyter.
1.86. PATTERSON, R.D., MILROY, R. & ALLERHAND, M. (1993). What is the octave of a harmonically rich note? In I. Cross & I. Deliege (Eds.), Music and the Cognitive Sciences 1993. (Proceedings of Cambridge Conference on Music and the Cognitive Sciences 1990) (pp. 69-81). Switzerland: Harwood Academic Publishers.
1.87. ROBINSON, A., HOLDSWORTH, J., PATTERSON, R. & Fallside, F. (1990). A comparison of preprocessors for the Cambridge recurrent-error-propagation-network speech recognition system. In Proceedings of First ICSLP, Kobe, Japan.
1.88. Watt, R.J., Bock, J., Thimbleby, H. & WILKINS, A.J. (1990). Visible aspects of text. In Proceedings of Conference on "Applying Visual Psychophysics to User Interface Design" (pp.309-325), Lavenham, May 1990, organised by British Telecom.
1.89. WILKINS, A.J. (1990). Visual display units versus visual computation. In Proceedings of Conference on "Applying Visual Psychophysics to User Interface Design" (pp. 257-265), (Lavenham, May 1990, organised by British Telecom).
1.90. WILKINS, A.J. (1991). Light right for sight: Health and efficiency in lighting practice. In E. Mills (Ed.), Proceedings of the 1st European Conference on Energy-Efficient Lighting (pp. 57-63). Stockholm: Swedish National Board for Ind. & Technical Development, Department of Energy Efficiency.
1.91. WILKINS, A.J. (1993). Possibilities for migraine therapy using coloured glasses? In T.J. Steiner & L.A.H. Hogenhuis (Eds.), Headache and Migraine (Proceedings of the Anglo-Dutch Migraine Meeting, Canterbury, 1992) (pp. 10-15). Utrecht: Bunge.
1.92. WILKINS, A.J. (1994). Reading and individual preferences for illuminant chromaticity. In Proceedings of Lux Europa (Edinburgh).
1.93. WING, A.M., ALLISON, S., Cooper, N. & Thompson, S. (1992). Cooperative timing on the River Cam. In C. Auxiette, C. Drake & C. Gerard (Eds.), Proceedings of the Fourth Workshop on Rhythm Perception and Production (pp. 85-95) Bourges, France, June 1992.
Technical Reports and Theses
1.94. ALLERHAND, M. & PATTERSON, R.D. (1993). Measurement of stress in speech. AAM HAP Progress Report No. 3, APU Contract Report.
1.95. ALLERHAND, M. & PATTERSON, R.D. (1994). Vocal agitation as a predictor of emotion and stress. AAM HAP Final Report, APU Contract Report.
1.96. ALLERHAND, M. & PATTERSON, R.D. (1994). User documentation for the AIM software package. AAM HAP Progress Report No. 5, APU Contract Report.
1.97. ALLERHAND, M., PATTERSON, R.D., ROBINSON, K. & Rice, P. (1990). Optimisation of the SVOS algorithm. SVOS Progress Report No. 4, APU Contract Report.
1.98. ALLERHAND, M., PATTERSON, R.D., ROBINSON, K. & Rice, P. (1991). Application of the SVOS algorithm. SVOS Progress Report No. 5, APU Contract Report.
1.99. CHRONICLE, E. (1993). Visual discomfort and visual dysfunction in migraine. Unpublished PhD Thesis, University of Cambridge.
1.100. HAGGARD, P. (1991). The coordination of human prehension. Unpublished PhD Thesis, University of Cambridge.
1.101. Irino, T & PATTERSON, R.D. (1994a) A computational theory of asymmetric intensity enhancement. around acoustic transients. Technical Report: ISRL-93-9, NTT , Japan.
1.102. Lower, M.C. & PATTERSON, R.D. (1991). Spectrum modification of the PWID output between free-field and at-the-ear measurements. Institute of Sound and Vibration Report No. AC693/4.
1.103. Lower, M.C. & PATTERSON, R.D. (1992). Acoustical specifications and test procedures for the British Rail ILWS PWID. Institute of Sound and Vibration Research, Report No. AC 693/6, May.
1.104. Lower, M.C. & PATTERSON, R.D. (1992). Implementation of warning sounds in the British Rail 'ILWS' prototype. Institute of Sound and Vibration Research, Report No. AC 693/5, May.
1.105. Lower, M.C. & PATTERSON, R.D. (in press). Acoustical measurements on PWID prototype A09. Institute of Sound and Vibration Report No. AC693/5.
1.106. Lower, M.C., PATTERSON, R.D., Patten, I.T. & MILROY, R. (1991). Design and installation of a sound system to aid passenger evacuation from aircraft. Institute of Sound and Vibration Research Report No. AC684/1.
1.107. PATTERSON, R., ALLERHAND, M. & AKEROYD, M. (1993). AAM HAP Progress Report No. 4, August 1993. MOD 3207 Contract Report.
1.108. PATTERSON, R.D. & DATTA J (1994). Extending the frequency range of existing warning sounds. AAM HAP Auditory Warnings: Progress Report No. 1, APU Contract Report..
1.109. PATTERSON, R.D. & HOLDSWORTH, J. (1990). An introduction to auditory sensation processing. AAM HAP Progress Report No. 1, APU Contract Report.
1.110. PATTERSON, R., NORRIS, D., & CUTLER, A. (Eds.) (1990). Auditory/Connectionist Techniques for Speech (ACTS) Periodic Progress Report No. 1, and associated Annexe. CEC: Brussels, June 1990.
1.111. PATTERSON, R.D., NORRIS, D. & CUTLER, A. (Eds.) (1991). Auditory/Connectionist Techniques for Speech. (ACTS) Periodic Progress Report No. 2 and associated Annexes. Brussels: CEC.
1.112. PATTERSON, R.D., NORRIS, D. & CUTLER, A. (1992). Auditory/Connectionist Techniques for Speech. (ACTS) Periodic Progress Report No, 3. Brussels: CEC, July.
1.113. ROBINSON, K. (1993). Studies in timbre and pitch. Unpublished PhD Thesis, University of Cambridge.
1.114. WILKINS, A.J. (1992). Colour Control: A Computer Program to check the Chromaticity of Lenses and Issue Guidance for their Use. Applied Psychology Unit.
1.115. WILKINS, A.J. (1992). Information about Televisions and Computer Displays for Patients with Photosensitive Epilepsy. The British Epilepsy Association.
1.116. WILKINS, A.J. (1993). A System for Precision Ophthalmic Tinting: Manual for the Intuitive Colorimeter. Cerium Visual Technologies, 27 pp.
1.117. WILKINS, A.J. (1993). Intuitive Overlays for Use by Teachers and Optometrists. Institute of Optometry Marketing.
1.118. WILKINS, A.J. (1993). Photosensitive Epilepsy Associated with Playing Computer Games. Report commissioned by the Department of Trade and Industry, 34 pp.
Tests and Patents
1.119. HOLDSWORTH, J.W. & PATTERSON, R.D. (1991). Analysis of waveforms. UK Patent No. GB 2-234-078-A (23.1.91). London: UK Patent Office.
1.120. PATTERSON, R.D. & HOLDSWORTH, J.W. (1990). Apparatus and methods for the generation of stabilised images from waveforms. World Intellectual Property Organization, WO 90/14656 (29 November 1990), 1990.
1.121. WILKINS, A.J. (1994). Apparatus for and a Method of Obtaining an Ophthalmic Tint. UK Patent No. 2,246,427 (4th May 1994). London: UK Patent Office.
1.122. Wilkinson, P.R. & WILKINS, A.J. (1991). Optical Devices for Reducing Photophobia. Patent No. GB 9000165 (4.1.91). London: UK Patent Office.
1.123. WING, A.M. & BARTON, J. (1992). SwayWeigh UK Patent App No 9212845.3 Filing date: 17/6/92 by MRC for Analogue balance.
REFERENCES TO OTHER WORK
Assmann, P. & Summerfield, Q. (1990). Modeling the perception of concurrent vowels: Vowels with different fundamental frequencies. Journal of the Acoustical Society of America, 88, 680-697.
Assmann, P. & Summerfield, Q. (1994). The contribution of waveform interactions to the perception of concurrent vowels. Journal of the Acoustical Society of America, 95, 471-484.
Assmann, P. & Summerfield, Q. (1989). Modeling the perception of concurrent vowels: Vowels with the same fundamental frequency. Journal of the Acoustical Society of America, 85, 327-338.
Carlyon, R.P. (1991). Discriminating between coherent and incoherent frequency modulation of complex tones. Journal of the Acoustical Society of America, 89, 329-340.
Carlyon, R.P. (1994a). Detecting pitch-pulse asynchronies and differences in fundamental frequency. Journal of the Acoustical Society of America, 95, 968-979.
Carlyon, R.P. (1994b). Further evidence against an across-frequency mechanism specific to the detection of FM incoherence between resolved frequency components. Journal of the Acoustical Society of America, 95, 949-961.
Carlyon, R.P., Demany, L. & Semal, C. (1992). Detection of across-frequency differences in fundamental frequency. Journal of the Acoustical Society of America, 91, 279-292.
Carlyon, R.P. & Shackleton, T.M. (in press). Comparing the fundamental frequencies of resolved and unresolved harmonics: Evidence for two pitch mechanisms? Journal of the Acoustical Society of America.
Chapman, S. (1968). Catching a baseball. American Journal of Physics, 36, 868-870.
Ciocca, V. & Bregman, A.S. (1987). Perceived continuity of gliding and steady-state tones through interrupting noise. Perception and Psychophysics, 42, 476-484.
Dienes, Z. & McLeod, P. (1993). How to catch a cricket ball. Perception, 22, 1427-1439.
Guiard, Y. (1987). Asymmetric division of labor in human skilled bimanual action: The kinematic chain as a model. Journal of Motor Behavior, 19, 486-517.
Hartmann, W.M., McAdams, S. & Smith, B.K. (1990). Hearing a mistuned harmonic in an otherwise periodic complex tone. Journal of the Acoustical Society of America, 88, 1712-1724.
Ivry, R.B., Keele, S.W. & Diener, H.C. (1988). Dissociation of the lateral and medial cerebellum in movement timing and movement execution. Experimental Brain Research, 73, 167-180.
Johannson, R.S. & Westling, G. (1990). Tactile afferent signals in the control of precision grip. In M. Jeannerod (Ed.), Attention and Performance XIII: Motor Representation and Control. Hillsdale, N.J.: Lawrence Erlbaum Associates.
Kabbash, P., MacKenzie, I.S. & Buxton, W. (1993). Human performance using computer input devices in the preferred and non-preferred hands. In Proceedings of InterCHI '93 (pp. 474-481).
Kelso, J.A. S., Putnam, C.A. & Goodman, D. (1983). On the space-time structure of human interlimb coordination. Quarterly Journal of Experimental Psychology, 35A, 347-375.
Lee, D.N. (1976). A theory of visual control of braking based on information about time to collision. Perception, 5, 437-459.
Lee, D.N., Reddish, P.E. & Rand, D. (1991). Aerial docking by hummingbirds. Naturwissenschaften, 78, 526-527.
Luria, A.R. (1966). Higher Cortical Functions in Man. London: Tavistock.
Massaro, D.W. (1975). Backward recognition masking. Journal of the Acoustical Society of America, 58, 1059-1065.
Meddis, R. & Hewitt, M. (1991). Virtual pitch and phase sensitivity studied using a computer model of the auditory periphery: Pitch identification. Journal of the Acoustical Society of America, 89, 2866-2882.
Meyer, D.E., Keith Smith, J.E., Kornblum, S., Abrams, R.A. & Wright, C.E. (1990). Speed-accuracy tradeoffs in aimed movements: Toward a theory of rapid voluntary action. In M. Jeannerod (Ed.), Attention and Performance XIII: Motor Representation and Control. Hillsdale, N.J.: Lawrence Erlbaum Associates.
Milner, A.D. & Goodale, M.A. (1993). Visual pathways to perception and action. In T.P. Hicks, S. Molotchnikoff & T. Ono (Eds.), Progress in Brain Research. Amsterdam: Elsevier.Tresilian, J.R. (1991). Empirical and theoretical issues in the perception of time to contact. Journal of Experimental Psychology: Human Perception & Performance, 17, 865-876.
Moore, B.C.J. & Ohgushi, K. (1993). Audibility of partials in inharmonic complex tones. Journal of the Acoustical Society of America, 93, 452-461.
Patterson, R.D. & Hirahara, T (1989). HMM speech recognition using DFT and auditory spectrograms. ATR Technical Report, Kyoto, Japan.
Scheffers, M.T.M. (1983). Sifting vowels: Auditory pitch analysis and sound segregation. Doctoral dissertation, University of Groningen, Netherlands.
Slaney, M. & Lyon, R.F. (1990). A perceptual pitch detector. In Proceedings of the International Conference on Acoustics, Speech, and Signal Processing (pp. 357-360).
Vorberg, D. & Hambuch, R. (1984). Timing of two-handed rhythmic performance. In J. Gibbon & L. Allan (Eds.), Timing and Time Perception. Annals of the NY Academy of Sciences, 423, 390-406.
Westling G & Johansson R.S. (1984). Factors influencing the force control during precision grip. Experimental Brain Research, 53, 277-284
Wilkins, A.J., Nimmo-Smith, I., Slater, A.I. & Bedocs, L. (1989). Fluorescent lighting, headaches and eyestrain. Lighting Research and Technology, 21, 11-18. (Revised version of APU 2279)
Wing, A. (1988). A comparison of the rate of pinch grip force increases and decreases in Parkinsonian Bradykinesia. Neuropsychologia, 26, 479-482.
Wing, A.M., Church, R.M. & Gentner, D.R. (1989). Variability in the timing of responses during repetitive tapping with alternate hands. Psychological Research, 51, 28-37.
Yamanishi, J., Kawato, M. & Suzuki, R. (1980). Two coupled oscillators as a model for the coordinated finger tapping by both hands. Biological Cybernetics, 37, 219-225.
Collaborations
Carlyon
Outside UK
Demany, Semal - Psychoacoustics, Bordeaux
Maylor
UK based
Quinn - Psychology, St Andrews
Simpson - Geriatrics, St George's Hospital, London
Patterson, R
UK based
Rood - Defence Research Agency, Farnborough
Fallside, Giguerre - Engineering, Cambridge
Moore, Baer - Psychology, Cambridge
Robinson - Institute of Hearing Research, Glasgow
Outside UK
Anderson, Gilkey - Wright-Patterson Air Force Base, Ohio
Yost, Fay - Parmley Hearing Institute, Chicago
Feth, Krishnamurthy - Speech and Engineering, Ohio State
Blauert, Bodden - Engineering Bochum
Berthomier, Schwartz - Institute Communication de Parole, Grenoble
Kawahara, Tokura - Advanced Telephonic Research, Kyoto
Sellen
Outside UK
Buxton - Computer Science, Toronto
Wilkins
UK based
Brown - David Lewis Centre for Epilepsy, Cheshire
Evans - Institute of Optometry, London
Fish - National Hospital, London
Kennard - Neurology, Charing Cross Hospital, London
Martin - William Westley CP School, Cambridge
Smith, Troscianko - Psychology, Bristol
Smith - Psychology, Bradford
Outside UK
Kasteleijn - Instituut voor Epilepsiebestreijding, Netherlands
Jaen - University of Tukaman
Wing
UK based
Burgess-Limerick, Clapp, Howick, Jenner, Kirker - Rehabilitation, Cambridge
Fraser, Turton - Occupational Therapy, Cambridge
Sagar - Neurology, Sheffield
van Vliet - Stroke Research, Nottingham
Virji-Babul - Physiotherapy, Cambridge
Woodburn - Engineering, Cambridge
Outside UK
Bisiacchi - Psychology, Padova
Flanagan - Movement Science, Columbia, NY
Lederman - Psychology, Kingston, Ontario
Richardson, Biomechanics, Orsay, Paris
Ulrich - Psychology, Konstanz
Vorberg - Psychology, Braunschweig