| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
The Journal of Neurophysiology Vol. 79 No. 3 March 1998, pp. 1481-1493
Copyright ©1998 by the American Physiological Society
1 Developmental Neurobiology and Centre for Visual Science, Research School of Biological Sciences, Australian National University, Canberra 2601, ACT, Australia; and 2 Department of Psychology, University College London, London WC1E 6BT, United Kingdom
 |
ABSTRACT |
|---|
 Abstract
 Introduction
Methods
Results
Discussion
References
|
|---|
Ibbotson, Michael R., Colin W. G. Clifford, and Richard F. Mark. Adaptation to visual motion in directional neurons of the nucleus of the optic tract. J. Neurophysiol. 79: 1481-1493, 1998. Extracellular recordings of action potentials were made from directional neurons in the nucleus of the optic tract (NOT) of the wallaby, Macropus eugenii, while stimulating with moving sine-wave gratings. When a grating was moved at a constant velocity in the preferred direction through a neuron's receptive field, the firing rate increased rapidly and then declined exponentially until reaching a steady-state level. The decline in response is called motion adaptation. The rate of adaptation increased as the temporal frequency of the drifting grating increased, up to the frequency that elicited the maximum firing rate. Beyond this frequency, the adaptation rate decreased. When the adapting grating's spatial frequency was varied, such that response magnitudes were significantly different, the maximum adaptation rate occurred at similar temporal frequencies. Hence the temporal frequency of the stimulus is a major parameter controlling the rate of adaptation. In most neurons, the temporal frequency response functions measured after adaptation were shifted to the right when compared with those obtained in the unadapted state. Further insight into the adaptation process was obtained by measuring the responses of the cells to grating displacements within one frame (10.23 ms). Such impulsive stimulus movements of less than a one-quarter cycle elicited a response that rose rapidly to a maximum and then declined exponentially to the spontaneous firing rate in several seconds. The level of adaptation was demonstrated by observing how the time constants of the exponentials varied as a function of the temporal frequency of a previously presented moving grating. When plotted as functions of adapting frequency, time constants formed a U-shaped curve. The shortest time constants occurred at similar temporal frequencies, regardless of changes in spatial frequency, even when the change in spatial frequency resulted in large differences in response magnitude during the adaptation period. The strongest adaptation occurred when the adapting stimulus moved in the neuron's preferred direction. Stimuli that moved in the antipreferred direction or flickered had an adapting influence on the responses to subsequent impulsive movements, but the effect was far smaller than that elicited by preferred direction adaptation. Adaptation in one region of the receptive field did not affect the responses elicited by subsequent stimulation in nonoverlapping regions of the field. Adaptation is a significant property of NOT neurons and probably acts to expand their temporal resolving power.
The present paper is concerned with motion adaptation in the movement processing system that forms the input to the nucleus of the optic tract (NOT) and the closely associated dorsal terminal nucleus (DTN) in the wallaby, Macropus eugenii. The two nuclei contain neurons with similar response properties (Simpson 1984 Animals and physiological preparation
The results presented are derived from experiments on 11 adult wallabies, M. Eugenii. All experimental procedures were approved by the animal experimentation committee of the Australian National University and follow guidelines set out by the National Health and Medical Research Council of Australia. Surgical procedures were identical to those outlined in Ibbotson et al. (1994) Recording and data analysis
Extracellular responses were amplified and passed through a window discriminator that produced 1.2-ms pulses. The output from the discriminator was fed into an analogue-to-digital converter which sampled at 1 kHz (Metrabyte Das-20). The pulse times were stored for off-line analysis and also were binned for immediate presentation of peristimulus time histograms (PSTH) on the computer screen before counting spikes in selected time windows. The PSTHs were smoothed using a Chebyshev type II, forward-backward recursive digital filter with a cut-off frequency of 100 Hz (Rabiner and Gold 1975 Visual stimulation
The stimuli were achromatic sine-wave gratings presented on a display monitor (CCID7551: Barco Industries) and were generated by a computer-controlled video display driver (AT Vista: True Vision Inc.). The refresh rate of the monitor was 97.75 Hz, and each frame contained 480 lines (512 pixels/line). The screen luminance was 65 cd/m2 and contrast could be varied with seven-bit resolution after gamma correction. The usual contrast used during experimentation was 0.3 unless stated in RESULTS. The gratings could be positioned at any orientation and moved back-and-forth along the spatial frequency vector. The screen subtended either 67° (horizontally) × 52° or 90° × 66° depending on the viewing distance. Stimuli could be presented on the full screen or within rectangular or circular apertures. When presented within apertures, the surround was always at the mean luminance of the stimulating grating.
In total, we have recorded from 150 NOT neurons in the wallaby of which 108 (72%) neurons were classed as slow cells. Experiments specifically designed to study motion adaptation were obtained from 36 neurons (25 slow cells, 11 fast cells).
Rate of adaptation
When a textured image moves in the preferred direction through the receptive field of an NOT neuron its firing rate increases rapidly to a peak response in 10-50 ms. As movement continues, the mean firing rate of the neuron begins to decline in an exponential manner until it reaches a steady-state firing rate in 3-10 s (Fig. 1). At very low drift rates, the cells can take longer to reach the initial peak firing rate but the response still declines once the peak has occurred (e.g., see Fig. 6D in Ibbotson and Mark 1996
Motion responses after adaptation
The experiments described in this section were repeated on 29 neurons (22 slow cells, 7 fast cells). Figure 4A shows the response of a fast cell in an unadapted state. The stimulus cycle consisted of a 2.6-s period in which the stimulus screen was a uniform grey at the mean luminosity of the subsequently presented grating (resting phase). After the resting phase a sinusoidal grating appeared (spatial frequency: 0.25 cycles/deg) and moved in the cells preferred direction at a drift rate of 3.04 Hz (test phase). The grating was present for 0.5 s. At the end of the test phase, the screen returned to a uniform grey (recovery phase). During the resting and recovery phases, the cell fired at its normal spontaneous rate (39 spikes/s). The mean firing rate during the test phase was 115 spikes/s. Figure 4B shows the response of the same neuron but this time the test phase was preceded by a 2.5-s adapting phase in which a 0.25 cycles/deg grating was moved in the preferred direction at 6.04 Hz and a 0.1-s pause during which the stimulus returned to a uniform grey before beginning the test sequence. After a period of adaptation, the response in the test phase was smaller (77 spikes/s) than it was during the control experiment. After a period of adaptation, the responsiveness of the cells quite rapidly recovered to the unadapted state. For this reason, measurements of the effects of adaptation were conducted very soon after a period of adaptation and most attention was given to the first 0.5 s of the test response. It is clear, for example, that the largest effect of motion adaptation on the cell shown in Fig. 4, A and B, occurs in the early stages of the test response. Fast cells often have a large onset transient at the beginning of their responses to motion. One effect of adaptation was to narrow the width of this initial burst of firing, as was the case for the cell in Fig. 4, A and B.
Temporal tuning after adaptation
This section examines whether the temporal tuning of the neurons changed after a period of motion adaptation. Experiments consisted of measuring temporal frequency response functions under control conditions and after the cells had been exposed to a period of adaptation. In all these experiments, the adapting and test stimuli were moved in the cell's preferred direction and the spatial frequency was held constant throughout all stimulus phases. These experiments were repeated on 15 neurons (8 slow cells, 7 fast cells). The solid line in Fig. 5A shows a temporal frequency response function from a fast neuron in the unadapted state. Incorporated into the same graph is the frequency response function derived from the same cell after it had been adapted to stimulus movement at 6.04 Hz. The overall magnitude of the adapted response function is lower than the unadapted function. This agrees with the observation that responses are attenuated after a period of adaptation (Fig. 4). Of more interest, however, is that the curve shifted to the right along the temporal frequency axis after adaptation. Figure 5B shows a temporal frequency response function for a slow neuron before adaptation and after adaptation to a grating moving at a temporal frequency of 0.6 Hz. The response function has shifted to the right after adaptation.
Responses to impulsive stimuli
If a stimulus grating is jumped by an amount equal to or less than quarter of a cycle in the preferred direction in a single frame (10.23 ms), the cells in the NOT elicit large responses (Ibbotson and Mark 1996
Recovery from adaptation
Three cells were tested to find out the time taken to return to the unadapted state after a period of motion adaptation. The stimulus cycle consisted of a 5-s adaptation phase in which a grating was moved in the preferred direction at a constant temporal frequency, followed by four impulsive jumps of the stimulus 1.3, 3.4, 5.5, and 7.6 s after the adapting motion stopped. This cycle was repeated 16 times to obtain average responses. The peristimulus time histogram from one neuron stimulated in this way is shown in Fig. 10A. The response of the neuron oscillates during the adaptation phase at the same frequency as the stimulus (0.76 Hz). During the adaptation phase, the mean firing rate of the cell declined and an exponential function was fitted to the response, the decay time constant being 1.6 s. The first impulsive stimulus jump elicited a large onset transient and a rapid decay to a firing rate close to the spontaneous level. The response elicited by each successive impulsive jump led to a slower response decay. The decay of the responses to impulsive stimuli were fitted with exponential functions (Fig. 10A).
Is motion adaptation directional?
To test the directionality of motion adaptation, we adapted four cells with stimuli moving in the preferred or antipreferred directions and to gratings of the same spatial frequency that were stationary but reversed their contrast at a frequency equal to the drift rate of the moving grating. The 6-s period of adaptation was followed by a 0.5-s pause where the grating was stationary, a one-frame movement of the grating by 0.25 cycles, and a 5-s recovery phase.
Is adaptation generated locally?
Two kinds of tests were conducted to discover whether adaptation in one region of the receptive field of an NOT neuron could generate an adaptive effect in another independent region of the field. The receptive fields of individual NOT neurons vary from 30 to >90° in the horizontal plane. We could not obtain quantitative measures of receptive field sizes >90° due to technical limitations, but it is possible that some neurons have receptive fields that are considerably larger than this value. Using gratings moved within the boundaries of narrow vertically oriented rectangular apertures, it is possible to elicit clear motion responses from 90% of NOT neurons with apertures that subtend 4° horizontally by 52° vertically. Approximately 10% of neurons will not elicit reliable responses unless the stimulus width is TEST 1 (REPEATED ON 3 CELLS).
Two narrow vertical apertures (A and B) were positioned in the center of the screen, which, in turn, covered a large portion of the cell's receptive field. The apertures subtended 52° (vertically) by 3° and were surrounded by uniform grey areas. The apertures were separated by a 4° grey stripe. Continuously moving gratings could be presented in both apertures. Three types of experiment were carried out. In the first experiment, a moving grating was presented in aperture A (or B) and moved in the cell's preferred direction for 1.5 s while the grating in aperture B (A) was stationary. These responses were the controls. In the second experiment, a stationary grating was presented in aperture A (B) while a moving grating was presented in aperture B (A) for 1.5 s. After the 1.5-s period, there was a short pause (0.3 s), where both gratings were stationary, followed by a 1.5-s test phase in which the grating was again moved in aperture B (A). In the third experiment, the grating in aperture B (A) was moved for 1.5 s while the grating in A (B) was stationary. After a 0.3-s pause, the grating in aperture A (B) was moved for 1.5 s while that in aperture B (A) remained stationary.
TEST 2 (REPEATED ON 4 CELLS).
The stimulus consisted of the same areas as in test 1 (see preceding text) except that the apertures were 5° wide and were separated by a 4° wide grey stripe. A relatively large aperture had to be used during these experiments because only small responses were induced by the necessarily small impulsive movements when the aperture was made smaller. Responses from a smaller aperture were not well fitted by exponentials. The experiment was conducted as follows. First, responses were elicited by an impulsive 1/16 cycle movement of the stimulus in either the left or right aperture (control). Second, the adaptive 6-s movement of the grating and the impulsive stimulus were presented in the same test region. Third, the grating pattern in the left (or right) region was moved for 6 s and then the stimulus in the other region was moved in an impulsive fashion. A response to the impulsive stimulus jump always was elicited by a 1/16 cycle movement of the grating in both the left and right apertures. Prior adaptation in the same stimulus region led to a reduction in the time constant of the subsequent response to the impulsive movement (Fig. 14, test number 3). Prior adaptation in the other stimulus region did not lead to any significant change in the time constant of the impulse response (Fig. 14, test number 2). The results demonstrate that adaptation in one region of the receptive field did not influence the response elicited in another spatially separate region of the receptive field.
Apparent motion stimulation
Experiments were conducted with a stimulus that consisted of 10 bar pairs spread across 27.5° of the stimulus screen. All bar pairs were identical and consisted of two vertical areas (area A and area B), which subtended 52° (vertically) by 0.7°. The two areas were touching each other at there inner edge, and the separation from neighboring bar pairs was 1.5°. Theoretically, the experiment could have been conducted with just one bar pair, but the responses were small. By having 10 bar pairs operating in synchrony, it was possible to elicit quite large responses from the NOT neurons. All areas of the screen surrounding the bar pairs were grey (45 cd/m2). Areas A and B could be either the same grey as the surround or have a brightness of 82 cd/m2. The stimulus was as follows: 1) the brightness in area A increased in a step-wise fashion while area B remained grey; 2) the brightness in area B increased while area A remained grey; 3) the brightness in area A increased in a step-wise fashion followed 0.5 s later by the same in area B; and 4) the brightness in area B increased in a step-wise fashion followed 0.5 s later by area A. These stimulus sequences gave us the responses elicited by an increase in brightness in areas A and B with no apparent motion (1 and 2), and the responses elicited by apparent preferred (3) and antipreferred (4) motion.
The responses of NOT neurons to a maintained motion stimulus are characterized by a rapid rise to a peak firing rate followed by an exponential decline to a steady-state level. Adaptation of this form has been observed in all the NOT neurons we have studied, suggesting that it is a significant property of the cells' responses. The rate of adaptation varies as a function of the temporal frequency of the adapting stimulus. The most rapid adaptation is observed to motion in the preferred direction at temporal frequencies around that which elicits the cell's peak response. The decline in response through adaptation corresponds to an attenuation of the cell's temporal frequency response function. In the majority of cells examined, this attenuation is accompanied by a shift in tuning to a higher preferred frequency, although the latter effect is small in the majority of cases.
Possible sites of adaptation
Ibbotson et al. (1994) Why adapt?
By measuring the responses of the NOT neurons to small impulsive image movements, we could assess the temporal resolving power of the system. When the response to an impulsive jump is far wider in time than the stimulus, as was always the case with the NOT neurons, the response width at half height is inversely proportional to the neuron's cut-off frequency (Srinivasan 1983 Comparison with other species
No systematic study of motion adaptation has been conducted previously on the directional neurons in the optokinetic systems of mammals, but data suggesting that adaptation occurs are available in the literature. For example, Klauer et al. (1990)
INTRODUCTION
Abstract
Introduction
Methods
Results
Discussion
References
) and appear to behave as one functional entity, which will be described throughout this paper as the NOT. Neurons in the NOT are known to have a role in generating optokinetic nystagmus (e.g., Collewijn 1975a,b; Hoffmann and Schoppmann 1981; Mustari and Fuchs 1990; Schiff et al. 1988; Simpson 1984; Soodak and Simpson 1988) and have certain physiological properties peculiar to their task. They respond preferentially to wide-field visual stimuli moving in a specific direction. The cells do not habituate to repeated stimulation, and they respond in a direction-selective manner throughout a wide range of image velocities. In the wallaby, two general classes of NOT neurons have been reported, those that are most sensitive to slowly moving images (slow cells) and those tuned to detect rapidly moving images (fast cells). It is likely that the two cell types are postsynaptic to different populations of motion detectors (Ibbotson et al. 1994).
were the first to show that the responses of directional motion-sensitive neurons decline gradually during prolonged motion stimulation. They suggested that this decline in response and the subsequent reduction in background activity that occurred after the stimulus stopped may be correlated with psychophysically measured motion aftereffects. Since that first observation, the responses of many directional motion-sensitive neurons in a wide range of species were found to decline during continuous stimulation, and this effect has been termed motion adaptation (e.g., cat: Maddess et al. 1988; Vautin and Berkley 1977; von der Heydt et al. 1978; flies: Maddess and Laughlin 1985; rabbit: Oyster et al. 1972). This paper shows that the neurons in the NOT of the wallaby clearly adapt to moving images and examines the parameters that control this adaptation.
). These impulsive stimulus movements elicit large responses from the neurons. In the case of the slow cells, which are mainly considered in this paper, the firing rates rise to a maximum within 10-50 ms and then decay over a longer period, 
5 s (Ibbotson and Mark 1996). The decay time constants of the responses to impulsive stimulus movements may be taken to represent the time constants of the temporal filters in the motion detectors (Srinivasan 1983; de Ruyter van Steveninck et al. 1986; Zaagman et al. 1983). Motion adaptation dramatically reduces the time constants of the motion processing system. This is discussed in terms of the potential improvement in the system's temporal resolving power.
; Clifford et al. 1997). To make this comparison meaningful, evidence is presented that, in combination with earlier results (Ibbotson et al. 1994), supports the claim that a Reichardt-type motion detector may be in operation in the wallaby optokinetic system. The comparison between the adaptive properties of the NOT neurons and the model outputs shows many similarities but also highlights some shortfalls in the predictive capacities of the models.
METHODS
Abstract
Introduction
Methods
Results
Discussion
References
and are summarized below. The day before experimentation, each animal was given 10 mg phytomenadione (vitamin K1, Konakin 10, Roche) and 100 mg tranexamic acid (Vasolamin, Ilium) intramuscularly to reduce bleeding and 10 mg/kg lincomycin and 20 mg/kg spectinomycin (Lincospectin, Upjohn) intramuscularly along with 10 ml of paraffin oil orally to reduce the effects of fermentation in the gut. On the day of experimentation, anesthesia was initiated by injection of 30 mg/kg ketamine (Parnell) and 1.5 mg/kg xylazine (Rompun, Bayer). Anesthesia was maintained during surgery by intravenous injections of thiopentone (Pentothal, Boehinger-Ingelheim). Each animal was intubated with a 4-mm Sheridan cannula and placed into a stereotaxic apparatus. The head was tilted down by 25° such that the palpebral fissures were horizontal. A craniotomy exposed the brain between 9 and 16 mm posterior to Bregma and between 1 and 8 mm lateral of the sagittal suture. After surgery, animals were paralyzed with 0.25 ml suxamethonium (50 mg/ml, Scoline, Galaxo), respired with a mixture of 75% N2O-25% O2, and infused intravenously with 5.6 ml/h of Hartmann's lactate solution carrying 1 mg·kg
1·h1 of pentobarbitone and 4.2 mg·kg1·h1 of suxamethonium.
). Once located, the electrode was moved rostrally ~1.0 mm and laterally ~1.0 mm to the region known to contain the NOT and/or the DTN (Ibbotson et al. 1994). The NOT and DTN are very close to each other in the wallaby and cannot be distinguished physiologically, as is the case in other species. The two nuclei will be referred to here collectively as the NOT (see Ibbotson et al. 1994 for more details). After a period of recording, the NOT location was marked with an electrolytic lesion. As a standard procedure, the action potential waveforms from all recorded cells were analyzed to confirm that they had characteristics typical of the soma rather than the axon (Bishop 1964).
). The filter served the purpose of removing high-frequency noise while not altering the phase of any frequency response components. Unit recordings were stable for 10 h of continuous recording.
). For the slow moving pattern described above, each frame was refreshed every 20.46 ms (2 frames), which is lower than the integration times for the majority of cells. Therefore, the cells were presented with smooth movement rather than saltatory motion. This latter characteristic was confirmed by Fourier analysis of the PSTHs derived from the recordings because no significant frequency components were observed at half the frame rate of the monitor (i.e., 48.875 Hz).
RESULTS
Abstract
Introduction
Methods
Results
Discussion
References
). These response properties were observed in all of the 150 neurons in our total sample. When stimulated through a relatively narrow aperture (width: 12°; height: 52°), a distinct response oscillation is superimposed on the mean firing rates of the neurons (Fig. 1A). The oscillatory component is composed of two frequency components, one at the drift frequency of the stimulus and the other at twice the frequency of the input. Higher frequency components do not usually add significantly to the responses (Ibbotson et al. 1994).
View larger version (35K):
[in a new window]
FIG. 1.
A: response of a fast neuron to a sinusoidal grating (spatial frequency: 0.25 cycles/deg; contrast: 0.3) moving in the cells preferred direction at 3.04 Hz. Pattern movement began at 0 on the abscissa and stopped 6.138 s later. Initial response is large and slowly declines during the period of motion. Spike rate of the neuron oscillates at a frequency that matches the temporal frequency of the moving grating. B: response of the same neuron to the grating moving at 18.24 Hz. Both peristimulus time histograms (PSTHs) are fitted with exponential functions.
View larger version (23K):
[in a new window]
FIG. 6.
Responses to impulsive stimuli (0.5° jump of the stimulus in the cell's preferred direction in 10.23 ms) for 2 neurons. A and C: responses of the 2 neurons with no prior adaptation. B and D: responses after 4 s of adaptation at 3.2 Hz. Fitted lines are exponentials.
where R is the fitted response at time t, C is the peak response, T is the time constant of the decaying exponential, and E is the fitted steady-state response of the neuron.
(1)
View larger version (28K):
[in a new window]
FIG. 2.
Time constants of exponentials fitted to the decline in response after motion onset as a function of the temporal frequency of the moving grating for 4 cells (A, C, E, and G). Temporal frequency response functions of the same neurons are presented below each of these graphs (B, D, F, and H). A temporal frequency response function plots the mean responses elicited by movement stimulation at a number of temporal frequencies when the spatial frequency is held constant. 
, results with a 1 cycles/deg grating; - - -, results with a 0.5 cycles/deg grating, respectively. Error bars are standard errors.
). When the spike rate was averaged over a 5-s time period, the maximum response shifted to lower temporal frequencies because the responses close to the optimum frequency decayed very rapidly, thus reducing the mean response amplitudes at those locations (Fig. 3, - - -). Throughout this paper, all temporal frequency response functions were obtained by measuring the mean initial firing rates of the cells.
View larger version (21K):
[in a new window]
FIG. 3.
Temporal frequency response functions from the same neuron but calculated from the initial (0.5 s) and first 5 s of the response. , temporal tuning obtained when the response was measured over the first 0.5 s; - - -, tuning after measuring the responses over the first 5 s. Error bars are standard deviations.
View larger version (27K):
[in a new window]
FIG. 4.
A: response of a fast neuron to the movement of a grating in the cell's preferred direction at 3.04 Hz. Duration of motion is shown by the top bar. B: response of the same neuron to the same test duration after a period of motion adaptation at 6.04 Hz. C: response of a slow neuron to a grating moving at 0.76 Hz. D: response of the same neuron to the same test stimulus after 0.9 s of adaptation at 0.38 Hz. Adapting and test periods are shown by the horizontal bars. PSTHs are the averages of 32 repetitions of the stimulus cycle.
View larger version (31K):
[in a new window]
FIG. 5.
A and B: temporal frequency response functions for 2 neurons extracted by measuring the spike rates in the 1st 0.5 s of the test phase before ( ) and after (- - -) a period of adaptation. Lines through the data points are smoothed splines. Prior adaptation was at 6.04 Hz (A) and 0.6 Hz (B). Error bars are standard deviations. C: locations of the peak responses in the adapted and unadapted conditions for 15 neurons. D: location of V50 for the same cells in the adapted and unadapted conditions. 
, slow cells; 
, fast cells. If the points are below the solid diagonal lines in C or D, it shows that the peak temporal tuning or the V50 have shifted to the right along the frequency axis.
). Most slow cells elicit responses that have a rapid increase in firing rate followed by a slow exponential decay with a time constant of between 0.5 and 3 s. Fast cells have more complicated biphasic responses to impulsive stimuli that rise to a peak very rapidly then decrease their firing rates to a level below the cell's spontaneous rate for ~70 ms. In the present paper, we have concentrated on the slow cells (n = 25) because they show simple exponential decays when stimulated with impulsive stimuli.
View larger version (27K):
[in a new window]
FIG. 7.
A: mean responses over the 1st 1.5 s for responses to impulsive movements plotted as functions of adapting temporal frequency. Two spatial frequency gratings were used in the adapting phase (1 cycles/deg, ; 0.5 cycles/deg, - - -) and the results from each grating are plotted as separate lines. B: temporal frequency response functions for the same neuron using the same spatial frequency gratings. Responses elicited by an impulsive image movement in the preferred (C) and the antipreferred directions (D).
) elicits the largest responses for all tested temporal frequencies. The time constants obtained after adaptation to the 0.5 cycles/deg grating are consistently smaller for all temporal frequencies than the time constants obtained after adaptation to the 2 cycles/deg grating. The adaptation appears to be tuned to the temporal frequency of the stimulus, but the absolute firing rates of the cells and/or the spatial frequency of the stimulus also appear to have a controlling influence on the degree of adaptation. Such behavior would be expected if adaptation were governed by the response rate of elementary motion detectors with separable spatiotemporal frequency responses, such as the Reichardt detector (Clifford et al. 1997).
View larger version (29K):
[in a new window]
FIG. 8.
A and C: decay time constants of responses to impulsive stimuli plotted as functions of the temporal frequency of the prior adapting grating for 2 cells. Four different spatial frequency gratings were used in the adapting phase, and each line shows the results from a different spatial frequency. B and D: temporal frequency response functions of the same neurons obtained using the same 4 spatial frequency gratings. Spatial frequencies were: 2 cycles/deg (···), 1 cycles/deg (- - -), 0.5 cycles/deg ( ), 0.25 cycles/deg (· - ·). Error bars are standard deviations.
View larger version (28K):
[in a new window]
FIG. 9.
Data from Fig. 8, A and B, have been replotted in normalized form on temporal frequency axes (A and B) and on velocity axes (C and D). Line conventions as in Fig. 8. Normalization was achieved as follows. In A and C, for each spatial frequency, the highest firing rate was set to 1 and the lowest to 0. In B and D, the functions were normalised relative to the largest response.
View larger version (26K):
[in a new window]
FIG. 10.
Recovery from adaptation. A: response of a slow cell to 5 s of continuous motion stimulation followed by 4 impulsive jumps of the stimulus. Fitted lines are exponentials. B: decay time constants of responses to impulsive stimuli as functions of the recovery time after motion stopped for 3 neurons. C: mean responses from the first 1.5 s of the responses to impulsive stimulation for 3 neurons. Decay time constant during continuous motion stimulation was 1.6 s.
View larger version (30K):
[in a new window]
FIG. 11.
Responses to impulsive stimuli for a neuron with no prior adaptation (A; control), after 5.5 s adaptation to preferred direction motion (B), adaptation to antipreferred motion (C), and adaptation to counterphase flicker (D). Peak responses were not changed by adaptation but the time constants of the impulse responses decreased, particularly when the image moved in the preferred direction. , fitted exponentials.
View larger version (11K):
[in a new window]
FIG. 12.
A: time constant of the decay of a cell's response to an impulsive stimulus under control conditions (C), after adaptation in the preferred direction (P), adaptation in the antipreferred direction (A-P), and adaptation to counter-phase flicker (F). B: mean of the normalized time constants of 4 cells. Error bars are standard errors.
8°. Some 40% of neurons tested elicited clear, reliable responses when the stimulus width was only 2° (also see Ibbotson et al. 1994). This is the minimum stimulus width that elicits a measurable response. The value of 2° does not necessarily correspond to the receptive field sizes of the elementary motion detectors that provide the motion input.
) and when aperture A was stimulated first and aperture B stimulated second (- - -). The response to stimulation in aperture A is not affected by stimulation in B. Figure 13B shows the response when the stimulus moved in aperture B () and when the grating moved in B after the grating had already moved in A (- - -). The response to stimulation in aperture B is not affected by stimulation in A. Figure 13C shows the responses elicited when aperture B was stimulated () and when B was stimulated after a period of adaptation in the same region (- - -). The response elicited by stimulation in aperture B after a period of adaptation in the same zone is clearly attenuated as compared with the control. Figure 13D shows the normalized responses from three cells tested in this way. The graph shows that adaptation and testing in the same region of visual space leads to the responses being attenuated (test number 3) while adapting and testing in different regions of visual space causes no significant change in response size (test number 2).
View larger version (28K):
[in a new window]
FIG. 13.
A: response of a neuron to motion in aperture A only ( ) and in aperture A followed by B (- - -). B: response to motion in aperture B only () and in aperture A followed by B (- - -). C: response to motion in aperture B only () and in B followed by the same region (- - -). D: normalized responses for 3 neurons tested as in A-C. Test number 1 is the control, test number 2 is after adaptation in a different area, and test number 3 is after adaptation in the same area. All data was normalized relative to the largest response. Error bars are standard deviations.
View larger version (13K):
[in a new window]
FIG. 14.
Time constants of a cell's response to an impulsive stimulus, test number 1, under control conditions, test number 2, after adaptation to motion in a different region of the cells receptive field to the test stimulus, and test number 3, after adaptation in the same region of the receptive field as the test. Error bars are standard deviations.
View larger version (22K):
[in a new window]
FIG. 15.
Responses to apparent motion stimuli. Small vertical lines show the times where the brightness steps began. A: response to step-wise increase in brightness in area A only. B: response to step-wise increase in brightness in area B. C: response to step-wise increases in brightness in area A followed 0.5 s later by area B. D: reverse of C.
DISCUSSION
Abstract
Introduction
Methods
Results
Discussion
References
). However, the question then remains whether adaptation occurs within the motion detectors or whether it reduces the temporal integration of the outputs of individual motion detectors before spatial integration of those responses in the NOT. This issue is discussed further in the context of a proposed model of motion adaptation.
have presented evidence suggesting that a Reichardt-like motion detector (Reichardt 1961) may be in operation in the pathway leading to the NOT in the wallaby. This evidence centers around the observation that the responses of the neurons oscillate when the stimulus is a moving sine-wave grating (Fig. 1). The oscillatory responses consist of two frequency components, one at the drift frequency of the stimulus and the other at twice the frequency of the input. Only under certain stimulus conditions are higher frequency components observed. The oscillatory responses suggest that a second-order nonlinearity is in operation in the motion detectors, as is the case in the Reichardt detector (Egelhaaf et al. 1989). Results presented in the present paper demonstrate that motion in the preferred direction is generated by a nonlinear excitatory mechanism (Fig. 15), which further supports the use of a Reichardt-like motion detector to model the wallaby system.
have elaborated the basic Reichardt model so that it adapts to moving stimuli and more accurately predicts the response characteristics of the NOT neurons. The elaborated model will be referred to as the adaptive Reichardt detector (ARD). In the ARD model, the response of each elementary motion detector governs the adaptation of the temporal delay filters within that detector. The strongest adaptation thus is observed for stimulation at the detector's preferred temporal frequency. The ARD makes several predictions consistent with the observed adaptive properties of NOT neurons: the rate of decay of the response to maintained motion increases with temporal frequency up to the peak temporal frequency tuning of the motion detectors, adaptation is strongest to motion in the preferred direction, the value of the adapted time constant is a U-shaped function of adapting temporal frequency with the shortest time constants around the preferred temporal frequency, and the minimum locus of adaptation is the receptive field of a single elementary motion detector.
). Thus halving the time constants in the ARD doubles the peak temporal tuning. However, the lateral shifts in the peak temporal tuning of the NOT cells were generally much smaller than would be predicted from the large changes in time constants. A possible explanation for this result is that further filtering occurs after motion detection. If we suppose that motion detection occurs in the retina, as suggested previously (Ibbotson et al. 1994), the second stage of filtering may occur in the inputs to the NOT neurons themselves. Low-pass temporal filters efferent to the elementary motion detectors would serve as temporal integrators, attenuating the high-frequency oscillatory components of the motion detector responses. When two filters are placed in series, the resultant temporal frequency response function is the product of the two constituent filters' temporal frequency response functions, whereas the resultant impulse response is the convolution of the two impulse responses. Thus the presence of additional temporal filters afferent to the NOT would distort the predicted inverse relationship between the temporal frequency tuning and time constant in the responses of NOT cells.
), i.e., neurons with broad responses to impulsive jumps have low cut-off frequencies and vice verse (e.g., Ibbotson and Mark 1996). After a period of motion adaptation, the responses to impulsive jumps always became narrower, as revealed by the reduced time constants. Narrower responses are a clear indication that the temporal resolution of the cells is enhanced after adaptation. The temporal resolution in this context is that of discriminating variations in slip velocity. Zaagman et al. (1983) modeled the outputs of two motion detector systems, one with filters that had fixed time constants and another with filters that had adaptive time constants. They showed that the adapting system resulted in a far better temporal representation of the moving image than was possible with the system with fixed time constants. The latter system blurred the image, and this blur increased with increasing image velocity. de Ruyter van Steveninck et al. (1986) suggested that adaptive temporal filters were of benefit for information processing in motion vision because they could optimize the performance of nonlinear interactions such as those that occur in motion detectors. This proposal suggests that some adaptive filtering should occur before the nonlinear interactions.
stimulated neurons in the ferret NOT with wide-field stimuli, which moved in the preferred direction for 5 s. They present a PSTH from a neuron that elicits a transient increase in firing rate soon after motion onset followed by a gradual decline in response over the next 3-4 s until it reaches a steady-state firing level. This decline in response is very similar to the adaptation observed in wallaby NOT neurons under similar stimulus conditions (e.g., Fig. 1). Pereira et al. (1994), using 6-s periods of constant velocity motion stimulation in the preferred direction, showed that the firing rates of neurons in the NOT of the opossum increased rapidly after motion onset then decayed slowly to a steady-state level over a period of ~5 s. The responses of directional retinal ganglion cells (RGCs), which are thought to form the input to the NOT in the rabbit, also were shown to decay exponentially during a period of continuous motion stimulation (Oyster et al. 1972). We have fitted the PSTHs presented by Oyster et al. with exponentials of the form used to fit the wallaby data, and these show that the rate of adaptation in the rabbit neurons increases with increasing image speed up to the peak velocity tuning for the cells. At higher velocities, the rate of adaptation declines, as occurs in the wallaby NOT neurons. Neurons in the basal optic root (nBOR) of the pigeon have been stimulated with sine-wave gratings that were moved for several seconds in the preferred direction (Wolf-Oberhollenzer and Kirschfeld 1990). As with the mammalian neurons described above, these cells show a transient increase in firing rate followed by a relatively slow response decay that leads to a steady-state level. The nBOR forms the main part of the accessory optic system in birds and performs a similar role to the NOT-DTN and other accessory optic nuclei in mammals (Simpson 1984). In the pigeon, it was found that the rate of decay of the transient component of the response (i.e., the rate of adaptation) was dependent on the temporal frequency of the stimulus grating, regardless of its spatial frequency (Wolf-Oberhollenzer and Kirschfeld 1994). This result corresponds with that found for the neurons in the NOT of the wallaby.
; Maddess and Laughlin 1985; butterfly: Maddess et al. 1991; bee: Ibbotson 1992). As with the neurons in the NOT of the wallaby, the adaptation was found to take place locally within the receptive fields of the insect neurons. Also, it was found that the adaptation process was governed by the temporal frequency of the adapting stimulus, at least at frequencies >2 Hz. Furthermore, a flickering stimulus was found to be sufficient to drive adaptation in the insect neurons even in the absence of any moving image (Borst and Egelhaaf 1987). In the insect system, a moving stimulus caused a greater change in the time constants of the motion detectors than flicker. However, the difference between the adaptive capacity of motion and flicker was less than in NOT neurons. The temporal frequency response functions of directional insect neurons were found to shift horizontally to the right and showed some response compression after a period of motion adaptation (Maddess and Laughlin 1985; Maddess et al. 1991). Clearly, the above results are similar to those seen in the wallaby, suggesting that similar mechanisms are in place in the optokinetic systems of mammals and insects (Clifford et al. 1997).
; Vautin and Berkley 1977). Maddess et al. (1988) showed that the degree of adaptation, as measured by the size of the response after 6.7 s of motion when compared with the initial response, was dependent on the temporal frequency of the stimulating grating. Saul and Cynader (1989) showed that responses elicited after a period of motion adaptation were reduced when compared with control values. These aftereffects were highly temporal frequency specific and bidirectional, i.e., adaptation in one direction caused frequency specific effects in both directions. Adaptation tended to shift a cell's preferred temporal frequency, the response being reduced mainly at frequencies close to the adapting value. This is a different result to that found in the NOT where the tuning functions were shifted consistently to the right. The results of experiments on cortical neurons suggest that there may be multiple inputs to the cortical cells and adaptation at a specific temporal frequency could selectively suppress just one channel. The NOT neurons, on the other hand, appear to receive their input from an array of elementary motion detectors distributed over a wide visual field, all with similar temporal tuning properties (Ibbotson et al. 1994). The temporal frequency dependency of adaptation is a common feature of NOT and cortical neurons. However, due to some differences between the observed properties it is not yet possible to suggest a common mechanism at the computational level.
| |
ACKNOWLEDGEMENTS |
|---|
Present address of C.W.G. Clifford: Brain and Vision Research Lab., Dept. of Biomedical Engineering, Boston University, 44 Cummington St, Boston, MA 02215.
| |
FOOTNOTES |
|---|
Address for reprint requests: M. R. Ibbotson, Developmental Neurobiology, Research School of Biological Sciences, Australian National University, GPO Box 475, Canberra 2601, ACT, Australia.
Received 30 September 1996; accepted in final form 25 November 1997 .
| |
REFERENCES |
|---|
|
Abstract
Introduction
Methods
Results
Discussion
References
|
|---|
This article has been cited by other articles:
|
|
 |
|
  N.S.C. Price, N. A. Crowder, M. A. Hietanen, and M. R. Ibbotson Neurons in V1, V2, and PMLS of Cat Cortex Are Speed Tuned But Not Acceleration Tuned: The Influence of Motion Adaptation J Neurophysiol, February 1, 2006; 95(2): 660 - 673. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. R. Ibbotson Contrast and Temporal Frequency-Related Adaptation in the Pretectal Nucleus of the Optic Tract J Neurophysiol, July 1, 2005; 94(1): 136 - 146. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. Cao, Y. Gu, and S.-R. Wang Visual Neurons in the Pigeon Brain Encode the Acceleration of Stimulus Motion J. Neurosci., September 1, 2004; 24(35): 7690 - 7698. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
N. S. C. Price and M. R. Ibbotson Direction-Selective Neurons in the Optokinetic System With Long-Lasting After-Responses J Neurophysiol, November 1, 2002; 88(5): 2224 - 2231. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. R. Ibbotson and N. S. C. Price Spatiotemporal Tuning of Directional Neurons in Mammalian and Avian Pretectum: A Comparison of Physiological Properties J Neurophysiol, November 1, 2001; 86(5): 2621 - 2624. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. S. Tolias, S. M. Smirnakis, M. A. Augath, T. Trinath, and N. K. Logothetis Motion Processing in the Macaque: Revisited with Functional Magnetic Resonance Imaging J. Neurosci., November 1, 2001; 21(21): 8594 - 8601. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. R. Ibbotson and C.W.G. Clifford Interactions Between ON and OFF Signals in Directional Motion Detectors Feeding the NOT of the Wallaby J Neurophysiol, August 1, 2001; 86(2): 997 - 1005. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. Kurtz, V. Durr, and M. Egelhaaf Dendritic Calcium Accumulation Associated With Direction-Selective Adaptation in Visual Motion-Sensitive Neurons In Vivo J Neurophysiol, October 1, 2000; 84(4): 1914 - 1923. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Visit Other APS Journals Online |
