INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The angular vestibuloocular reflex (AVOR) will alter its behavior to minimize image motion across the retina during head movements in response to natural phenomenon such as aging or illness as well as in response to experimental manipulations. Spectacles that either magnify or minimize the visual world will lead to changes in the AVOR that are in the appropriate direction to foster image stabilization on the retina (Miles and Eighmy 1980). For example, when tested in the dark with sinusoidal rotations of 0.5-2.0 Hz, monkeys that wore ×2 spectacles developed increased AVOR gains (1.5-1.8), and monkeys that wore ×0 minimizing spectacles developed decreased AVOR gains (0.2-0.3) (Lisberger 1984; Lisberger and Pavelko 1986). The complexities of AVOR adaptation in response to manipulations of the visual inputs have been described in numerous other studies (Bello et al. 1991; Gonshor and Melvill Jones 1976a,b; Khater et al. 1993; Melvill Jones and Davies 1976; Shelhamer et al. 1992; Yakushin et al. 2000).

Frequency selectivity has been demonstrated in the process that mediates AVOR adaptation. Monkeys that were rotated at a single frequency of either 0.2 or 2.0 Hz while wearing magnifying or minimizing spectacles and then tested over a range of frequencies (0.1-4 Hz) showed the greatest changes in gain at the adapting frequency (Lisberger et al. 1983). When monkeys were adapted using training frequencies of 0.5, 2, 5, 8, or 10 Hz, the greatest changes in gain and the greatest degree of frequency specificity were observed after adaptation with the lowest frequency stimuli (Raymond and Lisberger 1996). The gain changes after adaptation at the higher frequencies were less robust and less frequency specific.

We have recently described a velocity-dependent gain enhancement in the horizontal AVOR for rotational frequencies >2 Hz (Minor et al. 1999). This nonlinear behavior of the horizontal AVOR is also seen in the response to steps of acceleration, where the gain during the acceleration portion of the stimulus is significantly greater than that during the constant velocity portion of the stimulus. In addition, the relationship between head and eye velocity during the initial portion of an acceleration step is nonlinear, as it is best fit with a polynomial equation containing a cubic term rather than a linear equation. We have also shown that this nonlinear component of the response demonstrates greater modification than the linear component with adaptation to magnifying and minimizing spectacles (Clendaniel et al. 2001). After wearing magnifying spectacles for 1 wk, the squirrel monkeys not only demonstrated increases in gain, but there was a greater increase in the velocity-dependent gain enhancement for responses to sinusoidal rotations. In addition, during steps of acceleration, there was a greater increase in gain during the acceleration portion of the stimulus as compared with that during the constant velocity portion of the stimulus. After similar adaptation with minimizing spectacles, the velocity-dependent gain enhancement was abolished. Our mathematical model of the vestibular system, developed to explain the behavioral responses observed in normal monkeys and monkeys after unilateral labyrinthectomy or canal plugging, and after spectacle-induced adaptation, contains separate linear and nonlinear pathways, both of which contain separate adaptation processes. Due to the greater changes in the nonlinear component of the response after spectacle-induced adaptation, we hypothesized that the two components of the response could be adapted separately.

The purpose of the current experiment was to determine if the linear and nonlinear components of the response were under differential adaptive control. By using the frequency- and velocity-dependent nature of the nonlinear component, we were able to devise adaptation stimuli that would invoke only the linear component or both the linear and nonlinear components. If the neural signals that generate the linear and nonlinear components of the response combine prior to the adaptive element (1 central adaptation elementFig. 1A), then adaptation paradigms that only recruit the linear component should demonstrate adaptive changes in both the linear and nonlinear components of the response. If the neural signals that generate the linear and nonlinear components combine after the adaptive elements (2 adaptation elementsFig. 1B), then adaptation paradigms that only recruit the linear component should demonstrate adaptive changes solely in the linear component of the response. The results from the present experiment support the later hypothesis. We show that there is modification of the nonlinear component of the response only when the rotational stimuli used during the adaptation process recruit that portion of the response.

View larger version (11K): [in this window] [in a new window]   Fig. 1. Schematic representations of the possible relationships between the linear and nonlinear pathways and the adaptive elements. A: the signals from the linear and nonlinear pathways project to a shared central adaptation element. B: the signals from the linear and nonlinear pathways project to separate central adaptation elements.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Surgical procedures and eye-movement recording

The surgical and experimental procedures used for recording eye movements with the scleral search coil technique were identical to those that have previously been described for this laboratory (Minor et al. 1999). Each animal was seated in a plastic chair with its head restrained by securing the implanted bolt to a chair-mounted clamp. The chair was connected to a superstructure that was mounted to the top surface of a servo-controlled rotation table capable of generating a peak torque of 125 N-m (Acutronic, Pittsburgh, PA). The horizontal VOR was tested with the animal seated in the upright position in the superstructure and aligned such that the horizontal canals were in the earth-horizontal plane of rotation.

Rotational testing

The AVOR was measured in darkness pre- and postadaptation. The step stimuli consisted of 3,000°/s² accelerations to a peak velocity of 150°/s followed by a plateau of head velocity lasting 0.9-1.1 s and then deceleration at 3,000°/s² to rest. The direction, duration, and interstimulus interval were varied randomly from one trial to the next. Sinusoidal head rotations were 0.5-15 Hz, peak velocity 20-150°/s, and each stimulus frequency was given for 60 s. The order in which different frequencies and velocities were tested was varied.

Adaptation paradigms

To induce adaptation, monkeys were fit with minimizing (×0.45) spectacles that were attached to the acrylic skullcap. Adaptation was induced over 4 h with the minimizing spectacles combined with three different sum-of-sines stimuli. The high-frequencyhigh-velocity (HF-HV) stimulus had component signals of 2.3, 4.0, and 5.9 Hz, each with a peak velocity of 30°/s. As we will show, this stimulus contained frequencies and velocities sufficient to recruit both the linear and nonlinear components of the response. The high-frequencylow-velocity (HF-LV) stimulus had component signals with frequencies identical to the HF-HV paradigm, each with a peak velocity of 10°/s. While this stimulus contained frequencies of sufficient intensity, the velocities were below the threshold for the nonlinear component of the response. The low-frequencyhigh-velocity (LF-HV) stimulus had component signals of 0.3, 0.5, and 1.4 Hz, each with a peak velocity of 30°/s. This stimulus contained velocities of sufficient intensity, but the frequencies were below the level necessary to engage the nonlinear component of the response. Consequently, only the linear component was recruited during the HF-LV and LF-HV stimulus paradigms. After postadaptation testing, the spectacles were removed, and the monkeys were allowed to recover in their normal environment. There was 1 wk between adaptation sessions. Each animal underwent a different sequence of adaptation paradigms.

Data analysis

The data were analyzed off-line using software that we wrote in the Matlab (The Math Works) programming environment. The methods of analysis are similar to those that we have described previously (Lasker et al. 1999; Minor et al. 1999). Measurements of the gain of the VOR for the 3000°/s² to 150°/s steps were made during two components of the stimulus: the acceleration portion of the stimulus and after the plateau of head velocity had been reached. The acceleration gain of the VOR, G_A, was measured for each trial as the ratio of the slope of a line through the eye-velocity points to the slope of a line through the head-velocity points during a 20-ms period starting 20 ms after the onset of the stimulus. The velocity gain of the VOR, G_V, was measured from the ratio of the mean slow component eye and head velocity evaluated 100-300 ms after the plateau head velocity had been reached for each trial.

To determine the relationship between head and eye velocity during the initial portion of the step of acceleration, an average response to the steps of acceleration was obtained from 10 to 30 trials in each direction. First- through fifth-order polynomial fits were made to the head- and eye-velocity data extending from 10 to 40 ms after the onset of the stimulus. The order of the polynomial necessary and sufficient to account for the trajectory of the response was specified by the Bayesian information criterion (BIC) (Cullen et al. 1996; Galiana et al. 1995; Schwarz 1978). This analytic method takes into account the decrease in the difference between the fit and the data that will occur simply from the addition of high-order parameters to the model and weighs this decrease against the order of the model (Schwarz 1978). A reduction in BIC value justifies the use of a more complex (higher order) model, whereas an unchanged or increased value of BIC indicates that no additional information is obtained from an increase in the complexity of the model.

For the sinusoidal rotations, eye-position data were differentiated with a four-point central difference algorithm to obtain eye velocity. Saccades were removed through an interactive program from responses at frequencies <4 Hz, and an average cycle was obtained based on the data representing slow phase eye velocity at each point in time. Responses at frequencies 4 Hz were not desaccaded, and only cycles without saccades were included in the analysis. Successive cycles were averaged. The amplitude and phase of the response fundamental were obtained from a Fourier analysis as were the corresponding values for the head-velocity signal. Gains and phases for eye with respect to head velocity were expressed with the convention that a unity gain and zero phase imply a perfectly compensatory VOR. A negative phase indicates that eye movements lag head movements.

Statistical analysis

Results were described as means ± SD. Repeated-measures ANOVA was used to compare the data from more than two groups. Post hoc t-tests or one-way ANOVA tests were performed when the ANOVA demonstrated significant main effects or interactions, and the post hoc analyses are reported in RESULTS. Paired t-tests were used when preadaptation and postadaptation data in each animal were analyzed.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Responses to sinusoidal rotations

There was a significant reduction in gain after all three adaptation paradigms. During sinusoidal testing from 0.5 to 12 Hz (peak velocity of 20°/s), there was on average a 23.8% decrease in gain (Fig. 2A). The gain values at each testing frequency under a given adaptation paradigm were analyzed using a one-way ANOVA to determine if there were any frequency-specific effects. There was no significant effect of frequency under any adaptation paradigm (P values ranged from 0.59 to 0.95). In addition, post hoc analyses with Fisher's pairwise comparisons revealed no significant difference between any two frequencies. Thus the gain was uniform across testing frequencies, indicating that there was no frequency-specific adaptation as a result of the different stimulus parameters. There was no difference in the overall reduction in gain among the different adaptation paradigms when tested at 20°/s. Analysis of the responses to 4-Hz stimuli with peak velocities of 20, 50, 100, and 150°/s are depicted in Fig. 2B. After adaptation with the HF-LV stimulus, there was a marked reduction in gain, yet the velocity-dependent gain enhancement was still present (20°/s = 0.56 ± 0.12; 150°/s = 0.72 ± 0.16, P < 0.05). A similar reduction in gain, with preservation of the velocity-dependent gain enhancement, was noted postadaptation with the LF-HV stimulus (20°/s = 0.56 ± 0.09; 150°/s = 0.69 ± 0.09, P < 0.05). After adaptation with the HF-HV stimulus, there was both a decrease in gain and a loss of the velocity-dependent gain enhancement (20°/s = 0.52 ± 0.02; 150°/s = 0.54 ± 0.08, P > 0.1). These results indicate that the adaptation paradigms that only recruit the linear component of the response were sufficient to induce a decrease in gain but were not sufficient to cause adaptation of the nonlinear component. When adapted with stimuli sufficient to recruit both the linear and nonlinear components, there was a reduction in gain as well as a reduction in the velocity-dependent gain enhancement of the nonlinear component of the response. The response to adaptation with the HF-HV stimulus was similar to that seen after long-term (5-7 days) adaptation with ×0.45 minimizing spectacles (Clendaniel et al. 2001).

View larger version (13K): [in this window] [in a new window]   Fig. 2. Pre- and postadaptation gains to sinusoidal stimuli. A: gains to sinusoidal rotations from 0.5 to 12 Hz, peak velocity 20°/s.; HF-LV, gains after the high-frequencylow-velocity adaptation paradigm; LF-HV, gains after the low-frequencyhigh-velocity paradigm; and HF-HV, gains after the high-frequencyhigh-velocity adaptation paradigm. B: gains to 4-Hz stimuli with peak velocities of 20, 50, 100, and 150°/s. C: gains to 10-Hz stimuli with peak velocities of 20 and 100°/s. After adaptation with the ×0.45 spectacles, there were decreases in gain. The velocity-dependent gain enhancement was abolished only after the HF-HV adaptation paradigm.

Because one of the frequencies that composed the sum-of-sines stimulus during the high-frequency paradigms was 4 Hz, the difference in the 4-Hz test results could be a frequency-specific effect. This is an unlikely explanation for the observed behavior, however, as results similar to those at 4 Hz were observed when the VOR was measured at a frequency that was not used in the adaptation paradigm. The responses to sinusoidal rotations at 10 Hz (peak velocities of 20 and 100°/s) are depicted in Fig. 2C. There were significant reductions in gain with all adaptation paradigms. However, the velocity-dependent gain enhancement was present after adaptation with the HF-LV and LF-HV paradigms. After adaptation with the HF-LV paradigm, the mean gain at 20°/s was 0.53 ± 0.10 and the gain at 100°/s was 0.64 ± 0.15 (P < 0.1). The mean gain after adaptation with the LF-HV paradigm at 20°/s was 0.49 ± 0.07,and the gain at 100°/s was 0.57 ± 0.02 (P < 0.05). After adaptation with the HF-HV paradigm, there was a reduction in gain and an additional loss of the velocity-dependent gain enhancement, consistent with adaptation of both the linear and nonlinear components of the response (20°/s = 0.51 ± 0.02; 100°/s = 0.52 ± 0.06, P > 0.45).

Responses to steps of acceleration

There were significant reductions in response gain during both the acceleration (G_A) and constant velocity (G_V) portions of the 3,000°/s² to 150°/s steps of acceleration after each of the adaptation paradigms. Preadaptation the mean acceleration gain was 1.03 ± 0.19 and the velocity gain was 0.84 ± 0.08 for the three monkeys. After adaptation with the HF-LV paradigm, G_A was 0.90 ± 0.16 and G_V was 0.72 ± 0.08 (Fig. 3A). This represents 13 and 14% decreases in G_A and G_V, respectively. Postadaptation with the LF-HV paradigm (Fig. 3B), G_A was 0.87 ± 0.18 and G_V was 0.68 ± 0.07. This represents 15 and 19% decreases in G_A and G_V. Adaptation with the HF-HV paradigm induced a decrease in G_V and a greater decrease in G_A (Fig. 3C). Postadaptation G_A was 0.66 ± 0.13, a 36% decrease in gain, and G_V was 0.64 ± 0.08, a 24% decrease in gain. ANOVA was used to determine the statistical significance of the gain changes. Compared to the preadaptation condition, the LF-HV and HF-LV paradigms induced significant changes (P < 0.01) in gain and significant differences (P < 0.01) between G_A and G_V, but no significant interaction among G_A, G_V, and the adapted state. In both instances, the decrease in G_A was no different from the decrease in G_V, which is consistent with adaptation of the linear, but not the nonlinear, component of the response. In comparison to the preadaptation condition, the HF-HV paradigm induced significant differences in gain, between G_A and G_V, as well as a significant interaction among terms (P < 0.01 in all cases). The significant reduction in gain is consistent with adaptation of the linear component, and the significant interaction term is consistent with adaptation of the nonlinear component of the response. The summary data are plotted in Fig. 3D.

View larger version (33K): [in this window] [in a new window]   Fig. 3. Pre- and postadaptation responses to 3,000°/s2 to 150°/s acceleration steps. Dashed lines are the stimulus profile (inverted for ease of comparison). Average responses (±SD) are depicted by the white lines and shaded regions. A: there is a marked decrease in gain during both the acceleration and constant velocity portion of the stimulus postadaptation with the HF-LV condition. B: there is no difference in either GA or GV after the LF-HV and HF-LV conditions. C: there is a greater reduction in gain during the acceleration component of the stimulus after the HF-HV adaptation paradigm than after the HF-LV paradigm. D: summary graph demonstrates the parallel reduction in GA and GV after the HF-LV and LF-HV paradigms, and the greater reduction in GA after adaptation with the HF-HV paradigm. The asterisks indicate significant differences between GA and GV (P < 0.01).

A second method of demonstrating the nonlinear component of the AVOR, which has been used in previous investigations (Clendaniel et al. 2001; Lasker et al. 1999), is to analyze the relationship between eye and head velocity during the initial portion of the acceleration steps. As seen in earlier work describing responses to these stimuli in normal monkeys (Minor et al. 1999), a polynomial equation containing a cubic term provided a significantly better fit to the data than did either linear or second-order polynomial equations. The relationship between eye velocity and head velocity can be described by a third-order polynomial of the general form

(1)

where E_v is the eye velocity, H_v is head velocity, A, B, and C are the coefficients for the third-, second-, and first-order terms, respectively, and D is the intercept of the third-order fit on the eye-velocity axis (Minor et al. 1999). The initial VOR during the step of acceleration was evaluated with linear and polynomial fits relating eye to head velocity. The coefficients for the linear and cubic terms for the polynomial fits to the 3,000°/s² to 150°/s steps are presented in Table 1. Regardless of the adaptation paradigm employed, the trajectory of the response measured pre- and postadaptation was best fit by a polynomial equation containing a cubic term. After adaptation with the HF-HV paradigm, there was a consistent and relatively large reduction in the cubic term coefficient (Table 1). The changes in the cubic term coefficients after the HF-LV and LF-HV adaptation paradigms were less consistent and less robust. This observation was born out with statistical analyses. There was no difference among the coefficients for the cubic term in the preadaptation, HF-LV, and LF-HV conditions; after adaptation with the HF-HV paradigm, there was a significant reduction in the cubic term coefficient (paired t-test, P < 0.01). This finding is consistent with adaptation of the nonlinear component during the HF-HV adaptation paradigm. There were consistent but small reductions in the linear term coefficients after each adaptation paradigm. When analyzed using a paired t-test, there was no significant difference among the coefficients for the linear term in the HF-HV, HF-LV, and LF-HV conditions. When compared with preadaptation values, the only statistically significant reduction in the linear term coefficient occurred after the HF-HV adaptation (paired t-test, P < 0.01). This lack of statistical significance in the presence of consistent change is most likely due to variability in the linear term coefficients for the animals in this study.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Squirrel monkeys demonstrate a velocity-dependent gain enhancement during sinusoidal rotations >2 Hz, such that the gain of the AVOR increases with increasing velocity of rotation (Minor et al. 1999). We previously reported that after adaptation for 5-10 days with ×0.45 minimizing spectacles, squirrel monkeys demonstrate an overall decrease in gain and a loss of the velocity-dependent gain enhancement (Clendaniel et al. 2001). This loss of the velocity-dependent gain enhancement was hypothesized to represent adaptation of the nonlinear component response gain to a value of zero. This same decrease in gain and loss of the velocity-dependent gain enhancement is seen after short-term adaptation (4 h) with the HF-HV stimulus paradigm combined with the ×0.45 minimizing spectacles. A decrease in gain, with preservation of the gain enhancement at higher velocities, was seen after the HF-LV and LF-HV stimulus paradigms. Thus adaptation of the nonlinear component of the response is dependent on the characteristics of the head movements. Motion that is of insufficient frequency or velocity to recruit the nonlinear component, when paired with altered visual inputs, leads only to adaptation of the linear portion of the AVOR. If head-movement stimuli, sufficient to recruit both the linear and nonlinear components, are paired with altered visual inputs, then both the linear and nonlinear components of the AVOR will undergo adaptation.

It is unlikely that the observed results are a form of either frequency-specific adaptation (Lisberger et al. 1983; Raymond and Lisberger 1998) or context-specific adaptation (Shelhamer et al. 1992; Yakushin et al. 2000), where one would expect to see the effects of the adaptation only under conditions which mimic the adapting stimuli. First, the changes in gain were uniform across the tested frequencies (0.5-12 Hz, when tested with a peak velocity of 20°/s) regardless of the adaptation paradigm. This differs from previous reports of frequency-specific adaptation (Lisberger et al. 1983; Raymond and Lisberger 1998) and is most likely due to the fact that the adapting stimulus was a sum-of-sines stimulus composed of three component frequencies rather than one specific frequency. The variation in the component frequencies may have been sufficient to prevent a frequency-specific effect. Second, the results observed at 4 Hz (a component frequency of the adaptation stimulus in 2 of the 3 paradigms) were also seen at 10 Hz (a frequency not used in the adaptation paradigms). Third, one might argue that even though the frequencies and velocities of stimulation during the adaptation paradigms were different, if the accelerations of the HF-LV and LF-HV stimuli were similar and differed substantially from the HF-HV stimuli, then acceleration could be used as a contextual cue for the adaptation process. If this was the case, one might expect the observed postadaptation responses. However, for the HF-LV paradigm, the peak accelerations of the component frequencies were 144, 251, and 370°/s², and the peak acceleration during the sum-of-sines stimulus was 758°/s². The peak accelerations of the component frequencies for the LF-HV paradigm were 57, 94, and 264°/s², and the peak acceleration during the sum-of-sines stimulus was 405°/s². Although there was some overlap in the component accelerations between the LF-HV and the HF-LV stimuli, the accelerations in the two conditions were markedly different. Thus a form of context-specific adaptation, where acceleration is used as a context cue, cannot explain the postadaptation results.

Is there an anatomical basis to the linear and nonlinear components of the response? The model we have used to simulate the observed behavior in the horizontal AVOR in squirrel monkeys (with normal vestibular function, after unilateral plugging of the three semicircular canals, after unilateral labyrinthectomy, and after spectacle-induced adaptation) contains inputs to the reflex from what we term linear and nonlinear pathways (Clendaniel et al. 2001; Lasker et al. 1999, 2000; Minor et al. 1999). These signals are then combined centrally and project to the oculomotor neurons. There are several behavioral reasons for dividing the inputs into separate pathways and for suggesting that these pathways may be described by known anatomical components of the AVOR. The regularly discharging vestibular afferents display a constant rotational sensitivity with respect to frequency and velocity and are not driven into inhibitory cutoff for a wide range of stimulus intensities (Hullar and Minor 1999). This fits with the observed behavior of the AVOR across frequencies at 20°/s peak stimulus velocity (Lasker et al. 2000). The irregularly discharging afferents, on the other hand, display an increase in response gain with increasing frequency, are easily driven into inhibitory cutoff, and demonstrate an acceleration-deceleration asymmetry (Fernandez and Goldberg 1971; Goldberg and Fernandez 1971; Hullar et al. 2001). Thus the behavioral differences between the responses to 3,000°/s² acceleration and deceleration stimuli and the asymmetry between ipsi- and contralesional responses after unilateral canal plugging or labyrinthectomy are consistent with the firing patterns of the irregularly discharging afferents (Clendaniel et al. 2001; Lasker et al. 1999, 2000). Because the velocity-dependent gain enhancement is not seen in the response properties of either the regularly or irregularly discharging afferents, this aspect of the behavior must be conferred by central vestibular neurons.

It is possible that the linear and nonlinear components of the response are not separated at this early stage in the AVOR signal processing. It is conceivable that a signal representing head velocity is carried by the vestibular afferents to two subtypes of central vestibular neurons, one of which has greater nonlinear response dynamics than the other. The differences in the central vestibular neuron response dynamics could then account for the observed behaviors. Identification of the neuronal elements involved in these pathways will require data from single-unit recordings of the peripheral and central neurons involved in the control of the AVOR at stimuli similar to those employed in these studies and from experiments that selectively ablate the irregular afferents.

Regardless of the origin of the nonlinear component (peripheral or central), the results of the present experiment support the hypothesis that there are separate linear and nonlinear pathways controlling the horizontal AVOR and that there is differential adaptive control over the two pathways. One possible mechanism to explain the differential adaptive control of the linear and nonlinear pathways is that central vestibular neurons, presumably eye-head-velocity (EHV) cells/floccular target neurons (FTN), may receive a preponderance of either regularly or irregularly discharging vestibular afferents. The resulting changes in adaptation would then be dependent on which central neurons were active during the rotational stimuli. Another mechanism for the differential adaptive control could be the two classes of glutamate receptors (-amino-3-hydroxy-5-methyl-4-isoxazole-proppionic acid, AMPA, and N-methyl-D-aspartate, NMDA) found at the primary afferent-secondary vestibular neuron synapse and the interaction with -aminobutyric acid (GABA_B) receptors (Kinney et al. 1994; Peterson et al. 1996). If there were differences in the glutamate receptor subtypes between the regularly discharging afferent-secondary vestibular neuron synapse and the irregularly discharging afferent-secondary vestibular neuron synapse, then there would be potential for differential adaptation. Regardless of the actual cellular mechanism involved, to induce adaptation in the nonlinear component, the animal must experience head-movement stimuli, paired with altered visual inputs, that are of sufficient frequency and velocity to recruit the nonlinear component of the AVOR. If the animal is not exposed to stimuli that will recruit the nonlinear component of the response, then this component of the response will not adapt. Why would the animal need these separate linear and nonlinear components to control the AVOR? The squirrel monkey naturally makes head movements that are of very high acceleration, 2,000-10,000°/s², with peak velocities of 100-200°/s (Armand and Minor 2002). Because many of the movements are ballistic in nature rather than sustained, the linear pathway alone will not generate a response that is as compensatory as the combination of the linear and nonlinear pathways. Recall that in normal squirrel monkeys the gain during the constant velocity portion of the acceleration step (G_V) is less than one, and the gain during the acceleration portion of the acceleration step (G_A) is essentially equal to one (Minor et al. 1999). Consequently, the nonlinear component may have evolved to generate the required eye movements for these high-acceleration, high-frequency stimuli. Given the behavioral importance of this aspect of the response, it is reasonable to suppose that this component would be highly modifiable to optimize the animal's ability to maintain image stabilization on the retina.