| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Espace et Action, INSERM/Université Claude Bernard-Lyon, IFR19 Institut Fédératif des Neurosciences de Lyon, 69676 Bron cedex, France
 |
ABSTRACT |
|---|
 TOP ABSTRACT ResultsEffect of the double-step...Effect of the double-step...DiscussionRetention of saccade gain...Effects of adaptation on...Materials and MethodsREFERENCES |
|---|
; Shadmehr and Holcomb
1997) or submitted to visual perturbations
(Fernandez-Ruiz and Diaz 1999;
Tong and Flanagan 2003;
Caithness et al. 2004) or in
the case of the vestibulo-ocular reflex (VOR) in response to altered
visuo-vestibular interactions (Kuki et al.
2004). Adaptation of saccadic eye movements represents another
form of motor learning involved in the compensation for the physiological
(growing, aging, fatigue) or pathological changes affecting the oculomotor
system throughout life. Thus, when a saccadic eye movement becomes inaccurate
because it is too small (hypometric) or too large (hypermetric) relative to
the target position, saccadic adaptation mechanisms are elicited to gradually
adjust the saccade gain (ratio of saccade amplitude to target eccentricity)
such that the eyes again reach precisely the target. This phenomenon has been
clearly revealed in patients and in monkeys with weakened extraocular muscles
(Kommerell et al. 1976;
Abel et al. 1978;
Optican and Robinson 1980).
These three studies report that the saccade hypometria of the weakened eye was
progressively compensated for over successive days by an adaptive increase of
the saccade gain toward normal values. Saccadic adaptation can also be induced
in the laboratory in a noninvasive way by using the double-step target
protocol described by McLaughlin
(1967). This protocol consists
first in eliciting a saccade in response to the appearance of a visual target
and then, during the primary saccade, in displacing the target to create a
post-saccadic visual error, without conscious perception by the subject. The
repetition of such trials with either a backward or a forward intrasaccadic
target step leads to a progressive reduction of the post-saccadic error
through a decrease or an increase of the primary saccade amplitude,
respectively. The adaptation elicited in the double-step target paradigm is
faster than that observed under conditions of extraocular muscles weakness,
since it takes only a few hundred trials in human and 1000–3000 trials
in monkey. But in fact, as demonstrated by Scudder et al.
(1998) in the monkey, a
similar time course of adaptation is observed when the complexity of the set
of error signals is equated in the two paradigms. This type of results has led
to suggest that saccade adaptive modifications produced in the two different
paradigms share similar properties and neural substrate. Another advantage of
saccadic adaptation studied in the double-step target paradigm is that neither
the visual target perturbation nor the elicited corrective response
(corrective saccade) leads to conscious experience from the subject. Thus, the
behavioral changes induced in this adaptation protocol largely reflect the
true adaptive neural changes independently of any conscious strategic
response. This contrasts with the other types of motor adaptation reviewed
above in which subjects can use visual or proprioceptive feedback to
consciously detect the experimentally induced perturbation and may thus try to
deliberately adjust their motor responses.
Many studies have used the double-step target paradigm to investigate the
mechanisms underlying saccadic adaptation and have discovered a number of
functional and neurophysiological properties (for review, see
Hopp and Fuchs 2004). It has
been clearly shown in both man and monkey that adaptation of a single saccade
transfers to all saccades of the same direction and amplitude regardless of
their start and end positions in the orbit (vector specificity)
(Miller et al. 1981;
Deubel et al. 1986;
Deubel 1987;
Frens and Van Opstal 1994;
Albano 1996;
Straube et al. 1997;
Noto et al. 1999;
Watanabe et al. 2000). Thus,
adaptation of leftward saccadic vectors does not transfer to rightward
vectors, and vice versa. Saccadic adaptation also depends on the time when the
visual error signal becomes available after the primary saccade (temporal
specificity). Indeed, in both monkey and human
(Bahcall and Kowler 2000;
Shafer et al. 2000;
Fujita et al. 2002), varying
the time of occurrence of the second target step relative to the primary
saccade revealed that saccadic adaptation is maximal for a narrow
post-saccadic time window and declines sharply as the delay increases until
reaching a nonsignificant level for delays >1 sec. Furthermore, saccadic
adaptation in humans depends on the experimental tasks to elicit saccades
(saccade type specificity), as there is no or only a limited transfer of
adaptation between reactive saccades triggered by the sudden appearance of a
novel target and voluntary saccades directed to a predetermined target (e.g.,
scanning saccades). This suggests that the mechanisms and neural substrate of
adaptation differ for these two saccade types
(Erkelens and Hulleman, 1993;
Deubel
1995a,b;
Fujita et al. 2002). Finally,
although the neural substrate of saccadic adaptation remains to be completely
elucidated, there is a large body of evidence for an involvement of the
cerebellar vermis and fastigial nucleus in the adaptation of reactive
saccades, both in monkey (Optican and
Robinson 1980; Barash et al.
1999) and in human (Desmurget
et al. 1998; Straube et al.
2001).
Saccadic adaptation is supposed to elicit enduring changes in the neural
circuits involved in saccade generation. However, contrary to the long-term
storage of adaptive behavioral modifications that has been shown for the other
sensori-motor systems cited above, the retention of saccadic adaptation has
never been specifically investigated. Indeed, most studies of saccadic
adaptation have focused on the rate and total amount of gain changes observed
during the exposure phase, disregarding how long the effects of this initial
learning are kept in memory. The only cues available in the literature
concerning the retention of saccadic adaptation are indirect and based on
measures of saccade gain recovery during de-adaptation. It has been shown in
human (Deubel et al. 1986) and
in monkey (Straube et al.
1997) that the rate of de-adaptation in the presence of a visual
feedback provided by stationary visual targets is roughly as fast as that of
adaptation, suggesting that the retention of adaptation is very restricted.
However as envisioned by Deubel
(1995a,b),
because of the specificity of adaptive mechanisms relative to the different
saccade types in humans, the natural recovery from adaptation of reactive
saccades may be much longer than that evaluated in de-adaptation experiments
because subjects mostly perform voluntary saccades in their daily
activities.
Our study was thus aimed at measuring the duration of the after-effects
following the adaptation of reactive saccades, in other terms the retention of
adaptation over successive days. We induced a backward adaptation in five
human subjects by using the double-step target protocol. We measured the
saccade gain at different times immediately and up to 19 d after training. The
subjects underwent normal visual experience between the different measures.
Furthermore, we compared the efficiency and retention of adaptation of
rightward versus leftward saccades and investigated the effect of adaptation
on the saccade dynamics (main sequence) and latency. Preliminary results have
been reported in a short article (Alahyane
and Pélisson 2005).
|
Results |
|---|
|
TOPABSTRACTResultsEffect of the double-step...Effect of the double-step...DiscussionRetention of saccade gain...Effects of adaptation on...Materials and MethodsREFERENCES |
|---|
|
Effect of the double-step target protocol on the saccade gain |
|---|
|
TOPABSTRACTResultsEffect of the double-step...Effect of the double-step...DiscussionRetention of saccade gain...Effects of adaptation on...Materials and MethodsREFERENCES |
|---|
Main experiment
Figure 1A, shows the
relationship between the gain of saccades and the number of trials in the
adaptation session for subject B (day 0). The saccade gain progressively
decreased over trials as a consequence of the backward intrasaccadic step
introduced during the saccade. The gain decrease following the adaptive
learning was retained since the saccade gain was smaller in the post-test
performed after completion of the entire adaptation session (D0b), than the
gain in the pre-test. The other four subjects exhibited the same pattern of
results (data not shown). We plotted in
Figure 2 the saccade gain as a
function of the test sessions in each subject (A–E). We compared the
saccade gain between the different test sessions by individual analyses of
variance. We found a significant main effect of the test session factor in all
subjects (P < 0.001). The individual post-hoc LSD Fisher tests
indicated that all subjects exhibited a decrease in gain at D0a and D0b with
respect to the pre-test gain (P < 0.001). When the data from the
five subjects were pooled together (mean gain shown as a black trace in
Fig. 2), the one-way
repeated-measures ANOVA with "test session" as factor showed a
significant main effect (F(6,24) = 35.1; P <
0.001). From a nearly unity value in the pre-test (0.98 ± 0.03), the
gain had significantly decreased after the first half of the adaptation
session (D0a; post-hoc Fisher LSD tests; P < 0.001) and had
decreased further after the second half of the adaptation session (D0b),
reaching a significantly lower level than at D0pre (P < 0.001) and
than at D0a (P < 0.01). The percentage of gain change calculated
from these data revealed a mean gain reduction of -16.4% ± 4.6% at the
intermediate test session (D0a) and of -22.1% ± 4.7% at the end of
adaptation (D0b). Thus, these results indicate that the adaptation protocol
was successful in producing a significant backward adaptation of the
horizontal saccades with a larger after-effect reached after all adaptation
trials were completed.
|
|
We then studied how long the adaptive changes persisted over successive days after the adaptation session. The individual data illustrated in Figure 2 revealed that the persistence of the saccade hypometria was observed for all subjects, excluding subject D at D1 and excluding subjects A and D at D5. Concerning now the mean gain averaged across the five subjects, Figure 2 (upper left panel, black trace) shows that the mean gain began to increase as early as the following day, relative to D0b, but was still significantly smaller than the D0pre gain (post-hoc Fisher LSD tests; P < 0.001). At D5, saccades were still hypometric relative to those recorded before adaptation (P < 0.05). Thereafter, both at D11 and at D19, the gain returned to a level close to that obtained in pre-test (P = 0.11 and P = 0.47, respectively). In addition, the post-hoc analysis indicated that the gains at D5, D11, and D19 did not statistically differ from each other (P > 0.15). To quantify the level of maintenance of the adaptation induced at D0, we computed the amount of retention (see Materials and Methods). Figure 3 indicates that all subjects exhibited a relatively high amount of retention at D1, except subject D whose saccade gain was already back to a normal value, as already observed in Figure 2. On average, the amount of retention of adaptation reached 36% ± 17% at D1. It decreased to 19.7% ± 13.3% at D5 and reached a value of only 6.9% ± 10% at D19. Thus, the gain reduction resulting from the adaptation protocol decreased progressively over several days, being still significant 1 day after, and even 5 d after for three subjects.
|
), we
investigated if adaptation was better retained for saccades of large amplitude
than of small amplitude. We represented in
Figure 4A the average gain
(n = 5 subjects) of the two sizes of saccades as a function of the
test sessions. A repeated-measures ANOVA with two factors (target amplitude
x type of test session) revealed a significant effect of the test
session (P < 0.001; see previous paragraph) but no significant
effect of the amplitude factor (P = 0.148). Moreover, there was no
interaction between the two factors (F(6,24) = 0.272;
P = 0.94). In fact, the contrast analysis indicated that the
difference between the 10° and 20° amplitudes was only significant at
D0a (P < 0.05), with a gain change of -15.6% ± 5% for the
10° saccades and of -17.5% ± 5.1% for the 20° saccades. Thus,
as depicted in Figure 4A, the
statistically nonsignificant advantage of the 10° saccades gain relative
to the 20° saccades gain observed at D0pre was maintained throughout, and
the time course of adaptation-related gain change was nearly identical for the
two saccade amplitudes. In addition, the amount of retention of adaptation was
similar for the two sizes of saccades for all sessions
(Fig. 4B). In conclusion, the
amounts of adaptation obtained at D0 and retained over the testing period were
similar for saccades of 10° and 20°.
|
Control experiment
The control experiment produced results that clearly differed from those of
the main experiment. Figure 1B
illustrates the gain of all individual primary saccades performed by subject B
during the control experiment (day 0). Contrary to the adaptation condition,
subject B did not show in the pseudo-adaptation condition any change in gain
over trials or between pre-test and post-test. Note that the same pattern was
observed for the four other control subjects (data not shown). As in the main
experiment, we performed an ANOVA in each subject who was tested in the
control experiment to compare the saccade gain between the different test
sessions. Subjects G and H did not exhibit any significant main effect
(P > 0.6) contrary to subjects B, C, and F (P < 0.05).
In these three subjects, a post-hoc analysis was performed and revealed a
decrease of the gain at D1 and D5 for subject B (P < 0.001), a
decrease of the gain at D5 for subject C (P < 0.05), and an
increase of the gain at D5 for subject F (P < 0.01). But
importantly, none of the subjects presented a reduction of the gain at D0a and
D0b relative to the pre-test. The averaged data across the five subjects are
depicted in Figure 2
("mean", gray trace). The one-way repeated-measures ANOVA
comparing the different sessions did not reveal any significant main effect on
saccade gain (F(4,16) = 0.915; P = 0.479),
indicating that the saccade gain was equivalent between the different test
sessions. Thus, the results of the control experiment suggest that the
decrease of saccade gain observed in the main experiment was the consequence
of the double-step adaptation protocol and not of uncontrolled factors like
fatigue.
Effect of saccade direction
As both rightward and leftward saccades were exposed during our adaptation protocol, we had the opportunity to examine whether this adaptation affected the gain of rightward and leftward saccades to the same extent.
Time course of adaptation
The upper trace in Figure 5A
represents the time course of the mean saccade gain (n = 5) as a
function of the blocks of trials during the adaptation session, considering
rightward and leftward saccades separately. We can see that the gain evolved
in a similar way between the two types of saccades during A blocks. However,
beginning with the first B block, the gain of leftward saccades became smaller
than the gain of rightward ones. This difference was maintained over the
subsequent adaptation blocks. A two-way repeated-measures ANOVA with factors
"adaptation block" and "direction" revealed a
significant effect of the block factor (P < 0.001) as already
discussed, no effect of the direction factor (P = 0.23), but a
significant interaction (F(9,36) = 4.68; P <
0.001). The post-hoc analysis (Fisher LSD tests) indicated that only during
the two A blocks did the gain fail to exhibit a significant difference between
the two saccade directions (P > 0.11). Thus the gain changes of
rightward saccades and of leftward saccades were not identical, indicating
that the adaptation of leftward saccades was more efficient.
|
Saccade gain in test sessions
The relationship of the mean saccade gain (n = 5) as a function of
the test sessions is plotted for the two saccade directions separately in
Figure 5B. In the main
experiment (upper traces), the baseline gain (D0pre) was similar for rightward
and leftward saccades. But starting from the first post-test (D0a), the gain
of leftward saccades became smaller than that of rightward saccades, and this
difference was preserved over days. A two-way repeated-measures ANOVA (type of
test session x saccade direction) revealed a significant effect of the
test session factor (P < 0.001) as already described before, but
not of the "saccade direction" factor (P = 0.13).
However, there was a significant interaction between the two factors
(F(6,24) = 2.65; P < 0.05). Post-hoc Fisher
LSD tests indicated that the gain was equivalent between the two directions
only in the pre-test (P = 0.70), and a statistically significant
difference between directions was observed in all post-tests (P <
0.05). Furthermore, the leftward saccade gain was still significantly smaller
at D19 than the pre-test value (P < 0.05), whereas the rightward
saccade gain returned to a normal value as early as D5 (P = 0.98).
Thus, these averaged results suggest that the adaptation effect on leftward
saccades was stronger and more enduring. To see if this right-left asymmetry
was found in all subjects or only in some of them, we computed an index of
asymmetry (IA) as follows: {[rightward gain - leftward gain]/[0.5(rightward
gain + leftward gain)]} x 100. A positive IA indicated a higher gain for
the rightward direction than for the leftward one, whereas a negative value
indicated a superiority of the leftward saccade gain.
Figure 5C depicts, for the main
experiment, the IAs as a function of the test sessions for each subject and
for the average across the five subjects. In each subject, the saccade gain
was submitted to an ANOVA with test session and saccade direction as factors.
The five subjects showed a significant effect of the test session factor
(P < 0.001) as already described above, and of the saccade
direction factor (P < 0.01; except subject B, P = 0.08).
Subjects C, D, and E showed a significant interaction between the two factors
(P < 0.05) contrary to subjects A and B (P > 0.17).
Post-hoc Fisher LSD tests (subjects C, D, E) or contrast analyses (subjects A
and B) were then performed. The results are illustrated by the asterisks in
Figure 5C. Except subjects A
and B, the subjects exhibited a significantly higher gain for the rightward
saccades than for the leftward saccades as early as the D0a post-test: This
right superiority was not present in the D0pre session, was preserved over
days and tended to decrease at D19. Thus, these results suggest that, for
three subjects out of five, the effect of the double-step target protocol was
larger on leftward saccades, and this difference remained over successive
days.
Concerning now the control experiment, the difference of gain between the two saccade directions was much less pronounced and remained nearly constant across the successive test sessions. As shown in Figure 5B (lower traces), the gain averaged over the five subjects was slightly lower for leftward saccades than for rightward saccades, particularly at D0a and D0b. However, a two-way repeated-measures ANOVA (type of test session x saccade direction) did not reveal any significant effect of the test session (P = 0.43) and of the saccade direction (P = 0.20) factors or any significant interaction (F(4,16) = 0.32; P = 0.86). ANOVAs comparing the test sessions and the saccade directions were also performed in each subject. As already mentioned above, subjects G and H did not exhibit a significant effect of the test session factor (P > 0.62) contrary to the other three subjects (P < 0.05). Moreover, only subject G did not exhibit a significant effect of the saccade direction factor (P = 0.258). But importantly, none of the subjects showed any significant interaction (P > 0.23) except subject B (P < 0.05). The post-hoc Fisher LSD tests performed in subject B or the contrast analyses performed in the other four subjects indicated a superiority of the rightward saccade gain only in subjects C and F (see stars on the IAs plotted for individual subjects in Fig. 5D). Note however that this superiority of the rightward direction was explained by the asymmetry of gain already present in pre-test. In conclusion, since the between-session changes in the directional asymmetry were absent in the control experiment and followed in the main experiment a time course similar to the overall adaptation-related gain changes, we can propose that the gain asymmetry between rightward and leftward saccades was inherent to the adaptation phenomenon itself.
|
Effect of the double-step target paradigm on saccade dynamics and latency |
|---|
|
TOPABSTRACTResultsEffect of the double-step...Effect of the double-step...DiscussionRetention of saccade gain...Effects of adaptation on...Materials and MethodsREFERENCES |
|---|
We investigated whether the large saccade gain modifications induced by the adaptation protocol at D0 were accompanied by any change in saccade duration, saccade peak velocity, or both parameters. Figure 6 illustrates the duration (panel A) and peak velocity (panel B) of saccades recorded on day 0 before and after the adaptation session (black trace). This figure shows that saccade duration decreased after half of the adaptation session (D0a) and decreased further after completion of the adaptation session (D0b). A similar reduction was observed for the saccade peak velocity. A one-way repeated-measures ANOVA applied to duration revealed a significant main effect of the test session factor (F(2,8) = 17.4; P < 0.01). The post-hoc analysis (Fisher LSD tests) indicated that saccade duration in the pre-test (57 ± 5 msec) was significantly larger than that at D0a (53 ± 4 msec; P < 0.01) and than that at D0b (51 ± 3 msec); P < 0.001). In addition, the difference of duration between the D0a session and the D0b session failed to reach statistical significance (P = 0.072). The same ANOVA applied to the saccade peak velocity showed a significant effect of the test session (F(2,8) = 13.6; P < 0.01). The post-hoc analysis revealed that peak velocity was significantly larger in pre-test (396 ± 32 deg/sec) than in the two post-tests (D0a: 363 ± 29 deg/sec, P < 0.01; D0b: 352 ± 36 deg/sec, P < 0.01), and again, that saccade peak velocity did not significantly differ between the two post-tests (P = 0.214). Latency was also analyzed to determine any change related to the adaptation protocol or to factors such as anticipation or fatigue. The one-way ANOVA showed no effect of test session on latency (mean: 263 msec, F(2,8) = 1.15; P = 0.362).
|
In the control experiment, saccade duration and peak velocity were also submitted to a one-way repeated-measures ANOVA with test session as factor (Fig. 6, gray trace). Concerning saccade duration, there was a significant main effect (F(2,8) = 7.32; P < 0.05). The post-hoc analysis revealed that the duration at D0a (57 ± 6 msec) was similar to that at D0pre (56 ± 5 msec; P = 0.232) but that the duration at D0b (59 ± 6 msec) was significantly longer than that at D0pre (P < 0.01). In addition, the duration was larger at D0b than at D0a (P < 0.05). Concerning saccade peak velocity, no significant effect was observed (mean: 390°/sec; F(2,8) = 1.18; P = 0.355). As in the case of the main experiment, saccade latency was not modified (mean: 237 msec; F(2,8) = 0.13; P = 0.882). Thus the decrease in duration and peak velocity of saccades observed in the main experiment resulted from the adaptation-related decrease of saccade gain and not from fatigue.
Main sequence relationship
It is important to determine if these reductions of duration and peak
velocity were merely related to the decrease in saccade gain or if they
revealed an effect of the adaptation protocol on saccade dynamics. To resolve
this question, we analyzed the main sequence relationships.
Figure 7 plots the duration
versus amplitude relationships for the two subjects who were tested both in
the main experiment and in the control experiment (panels A and B,
respectively), for all saccades recorded in the D0pre and D0b sessions. We
fitted the relations obtained in the two sessions with a linear function (see
Becker 1989). The qualitative
analysis of these fits indicates that there was an increase of duration of
saccades recorded in post-test with respect to the pre-test data for subject B
in both the main and control experiments (top panels). In subject C, however
(bottom panels), there was no change in the duration versus amplitude
relationship between D0pre and D0b. These two subjects represented the two
extreme examples of the spectrum of behavior seen in the eight subjects. We
then quantitatively analyzed the relation between saccade duration and
amplitude in the whole group of subjects. This relation was first fitted by a
linear function in the pre-test session, separately for each subject and for
each saccade direction. Then, for each data point collected in the D0b
post-test session, we calculated a residual error representing the difference
between the actual saccade duration and the duration predicted from the
pre-test regression line. Then, the post-test residual errors were averaged
separately for each subject and for each saccade direction. The 20 mean
residual errors (two groups of five subjects x two saccade directions)
were submitted to a repeated-measured ANOVA with saccade direction as
within-subjects factor and group (main experiment versus control experiment)
as between-subjects factor. This statistical analysis did not demonstrate any
main effect of the saccade direction factor (P = 0.43) or of the
group factor (P = 0.53) or any significant interaction
(F(1,8) = 0.26; P = 0.62). The data were thus
pooled together across the two groups of subjects and the two saccade
directions. The overall mean residual error was significantly different from
zero (1.6 ± 2.4 ms, n = 20; t-test, P <
0.01), corresponding to a very small (2.8%) increase of the mean saccade
duration in the D0b post-test relative to the pre-test. Overall, these
analyses indicated that the very slight change in main sequence observed in
the post-test relative to the pre-test could not be related to the adaptation
phenomenon itself.
|
The same analyses were performed for the peak velocity versus amplitude
relationships, using this time an exponential fit of the pre-test data (see
Becker 1989) to compute the
post-test residual errors (data not shown). The ANOVA applied to the peak
velocity residual error revealed no significant effect of the saccade
direction or of the group factors (P = 0.38 and P = 0.48,
respectively), or any significant interaction (F(1,8) =
1.83; P = 0.21). The overall mean residual error was negligible
(-0.044 ± 0.24°/sec, corresponding to a 0.011% decrease of the mean
saccade peak velocity in the D0b post-test relative to the pre-test) and did
not differ from zero (n = 20; t-test, P = 0.43). In
conclusion, these peak velocity analyses reinforced the main conclusion of the
duration analyses above: The slight increase of saccade duration possibly
reflected a moderate development of fatigue, but no change in the saccade
dynamic properties could be specifically related to adaptation.
|
Discussion |
|---|
|
TOPABSTRACTResultsEffect of the double-step...Effect of the double-step...DiscussionRetention of saccade gain...Effects of adaptation on...Materials and MethodsREFERENCES |
|---|
|
Retention of saccade gain changes |
|---|
|
TOPABSTRACTResultsEffect of the double-step...Effect of the double-step...DiscussionRetention of saccade gain...Effects of adaptation on...Materials and MethodsREFERENCES |
|---|
) or after
a 24-h period under normal illumination conditions
(Noto et al. 1999). But to our
knowledge, our study is the first one to show in humans a retention of saccade
adaptation after a recovery period of several days under permanent visual
feedback conditions.
It may appear surprising that the adaptive gain changes can persist for
such a long time period. Recall that during this delay period, subjects were
free to perform their daily activities and were submitted to visual feedback
arising from the several thousands of saccades they performed each day. Deubel
et al. (1986) observed that
when human subjects performed saccades to single-step targets following
completion of a double-step target adaptation protocol, de-adaptation (i.e.
recovery from adaptation) proceeded with a time course similar to that of the
acquisition of adaptation. It can thus be asked why visual feedback failed to
completely de-adapt our subjects and to normalize their saccade gain. A likely
explanation is that the saccades performed by our subjects everyday differed
from those recorded during the test sessions by several aspects. One possibly
predominant aspect is the amplitude of the two types of saccades. Indeed,
Bahill et al. (1975) have
indicated that the amplitude of the majority of saccades generated
spontaneously falls well below 15°. Since saccadic adaptation is amplitude
specific (Semmlow et al. 1989;
Frens and Van Opstal 1994),
saccades generated by our subjects in between the recording sessions may have
been too small to allow the 10° and 20° saccades to fully recover from
adaptation, especially for the 20° saccades. In fact, this assumption can
be rejected since we found no difference of retention between these saccades
and 10° saccades, suggesting that the adaptation of larger saccades was
not better retained. Another factor that may act as a context-cue for saccadic
adaptation and that could more likely explain the maintenance of adaptation is
the type of saccade initiation: Saccades performed in everyday life are mainly
voluntary saccades directed to pre-determined targets, whereas those studied
in our protocol are reactive saccades to suddenly presented targets. Indeed,
the partial retention of adaptation of reactive saccades found in our study is
consistent with the known specificity of adaptation with respect to the type
of saccade initiation in humans (Erkelens
and Hulleman 1993; Deubel
1995a,b;
Fujita et al. 2002; N.
Alahyane and D. Pélisson, unpubl.). A context-specificity of adaptation
has been further demonstrated recently by Gaveau et al.
(2005). These investigators
showed that the adaptation of reactive saccades vanished over a 15-min period
of presentation of single target steps when subjects performed saccades of the
same type (i.e., reactive) but not when they performed voluntary saccades.
Finally, several other factors may act as context-cues for saccadic adaptation
and may have thus contributed to the long-term persistence of memory for the
reactive saccades in our study, including eye position
(Alahyane and Pélisson
2004), head position
(Shelhamer and Clendaniel
2002), or target distance
(Chaturvedi and Van Gisbergen
1997). Which of these factors (including the type of saccade
initiation), or combination thereof, contributed most to the observed slowness
of de-adaptation cannot be answered from our study. In any case, our study
demonstrates in human subjects that when the conditions for de-adaptation are
not optimal, adaptive changes of reactive saccades can be maintained in
long-term memory.
An unexpected result was that adaptation and its memory were stronger for
leftward saccades than for rightward saccades, especially for three out of
five subjects. Note that no such difference of amplitude between the two
directions was observed in the pre-test session at the start of the
experiment, ruling out any baseline difference of gain unrelated to the
adaptive process itself. Moreover, the appearance of this difference during
the third block of adaptation trials (B block) cannot be easily related to the
design of our protocol because (1) the two saccade directions were equally
represented in each block, including B blocks; (2) the same number of leftward
and rightward trials was performed in each hemifield for all blocks; and (3)
in the control experiment, three subjects did not show any difference in gain
between leftward and rightward saccades, and the rightward asymmetry found in
the remaining two subjects corresponded to a gain difference as early as the
pre-test session. In fact, we can only speculate that a minimum number of
saccades (>80) and/or a minimum amplitude of error signal (target step-2 =
33% relative to step-1 in B blocks) is required to allow the adaptation
process to fully develop and to reveal differences between saccade directions.
This difference of robustness of adaptation as a function of the saccade
direction was maintained over the long term: Whereas rightward saccades were
fully de-adapted as early as D5, leftward saccades remained hypometric until
the end of the test period, i.e., after 19 days. In fact, the much longer
retention of adaptation of leftward saccades resulted, on the one hand, from
the larger gain change observed during the adaptation session and, on the
other hand, from the weaker de-adaptation during the delay period. Such
asymmetry of gain changes is a much unexpected finding, which to our knowledge
has never been reported before. It is well known that an adaptation of a
single saccade direction does not affect saccades of the opposite direction in
human (Miller et al. 1981;
Deubel et al. 1986). This
suggests that the adaptation of leftward and rightward saccades relies on
separate, and most likely lateralized, neural circuits. Our study further
suggests that the neural control of adaptation of leftward saccades is more
efficient than that involved in adaptation of rightward saccades. As our
trained subjects were all right-handed, it is tempting to propose that the
adaptive control of saccadic eye movements may show a stronger contribution of
the nondominant hemisphere. This superiority of the right hemisphere has also
been invoked, e.g., for the control of attention
(Heilman et al. 1985;
Coull and Nobre 1998) and of
vestibular function (Dieterich et al.
2003). The asymmetry of recovery to a normal saccade gain (i.e.,
de-adaptation) over the long term may also be explained by a larger efficiency
of the nondominant hemisphere in the maintenance of the saccadic motor memory.
Another possible reason may be that subjects performed fewer leftward compared
with rightward saccades between the recording sessions during tasks such as
reading, thus slowing down de-adaptation of saccades in the leftward
direction.
|
Effects of adaptation on saccade dynamics |
|---|
|
TOPABSTRACTResultsEffect of the double-step...Effect of the double-step...DiscussionRetention of saccade gain...Effects of adaptation on...Materials and MethodsREFERENCES |
|---|
) found a decrease of
both peak velocity and duration without change in the main sequence. In
contrast, Abrams et al. (1992)
found a change in the main sequence associated with a selective decrease of
peak velocity in a backward target step paradigm (their first experiment). One
possible explanation of the absence of a decrease of duration in this latter
study may be that the adaptation session contained many fewer trials than in
our protocol. Straube and Deubel
(1995) demonstrated that the
forward target step paradigm led to a modification of the dynamics of a
"standard 12° saccade" (decrease in peak velocity and increase
in duration) during the adaptation session, whereas the gain decrease paradigm
led to a decrease of the peak acceleration/peak deceleration ratio, by
comparison with a control condition without intrasaccadic target step.
Concerning now adaptation in the monkey, Straube et al.
(1997) revealed no average
change in saccade dynamics after adaptation but observed a large variability
between animals, with the more adapted one exhibiting a slowing of saccade
dynamics with an increase of duration and a decrease of peak velocity. Note
that cautions have to be taken in the interpretation of the data from the
literature since the experiments differ in terms of the number of saccades
during the adaptation session, of the amplitude of target steps, or in the
choice of the sessions used to compare the saccade dynamical parameters
(pre-test, post-test, adaptation, or control session without intrasaccadic
step). In our study, the number of saccades and the size of intrasaccadic
target step were sufficient to elicit a fairly high level of adaptation, and
the post- versus pre-test comparison allowed us to minimize effects that were
not specifically related to sensori-motor adaptation. The lack of change of
saccade dynamical properties observed under these conditions suggests that the
plastic neuronal modifications underlying the backward adaptation of reactive
saccades do not include the brainstem saccadic pulse generator. Combined with
our previous proposal of a collicular or subcollicular locus for saccade
adaptation (see discussion in Alahyane et
al. 2004), the present work leads us to tentatively suggest that
the backward adaptive mechanisms modify a desired eye displacement signal at
the collicular level or at the level of reticular formation neurons that
transmit collicular signal to the pulse generator.
In conclusion, similar to numerous sensori-motor systems such as
visuo-manual tracking (Shadmehr and
Holcomb 1997), eyelid conditioning
(Medina et al. 2001), or VOR
(see for review Boyden et al.
2004), we found, for the saccadic system, that changes of saccade
amplitude that quickly develop during a double-step target adaptation protocol
can remain in memory for a much longer period of time. Studies of the learning
of the former sensori-motor responses showed that short-term changes in motor
behavior and their long-term storage may rely on different neural circuits
within or outside the cerebellum. By demonstrating the existence of long-term
changes of saccadic performance following adaptation, our study raises the
possibility to test in the saccadic system such dissociations between neural
substrates of the short-term development and of the long-term retention of
sensori-motor adaptation.
|
Materials and Methods |
|---|
|
TOPABSTRACTResultsEffect of the double-step...Effect of the double-step...DiscussionRetention of saccade gain...Effects of adaptation on...Materials and MethodsREFERENCES |
|---|
This study was conducted in eight voluntary subjects (age range: 20–45 yr) after getting their informed consent. All subjects had a normal or a corrected-to-normal vision. Five subjects (A, B, C, D, E) participated in the main experiment. Subjects B and C also performed the control experiment, together with three other subjects (F, G, H). Only subjects B and C were not naive about the purpose of this study. The study complied with the declaration of Helsinki.
Apparatus
Subjects were seated, in a dimly lighted room, in front of a concave spherical board containing red light-emitting diodes (LEDs; diameter, 3 mm) used as visual targets. The center of the board was aligned with the subject's naso-occipital axis at 1.10 m (sphere radius) from the cyclopean eye. The head was stabilized by means of a chin rest. All targets were located along the azimuth at eye level (0° axis), and their position ranged from -27.5° eccentricity (to the left of the center of the board) to 27.5° (to the right), separated every 2.5°. Two other LEDs used for calibration were situated at ±12.5° on the vertical axis. Subjects were required to carefully follow the targets with their eyes.
Protocol in main experiment
Adaptation session
Saccadic adaptation was induced by the double-step target protocol. At the
beginning of a trial, a fixation point was illuminated at a random position on
the 0° axis for a randomly alternating period (1600 msec or 2000 msec).
Then, the fixation point was turned off while a peripheral target (target-1)
appeared randomly to the right or to the left (step-1). As soon as the eye
velocity exceeded a 50°/sec threshold during the saccadic response toward
target-1, the target was displaced again (step-2) to a backward position
(target-2). Step-2 was backward relative to step-1 in order to induce a
decrease of the saccade amplitude. Each final position of the target in a
double-step trial (target-2) became the starting point (fixation point) of the
following trial. The adaptation session was composed of three different types
of blocks (A, B, and C). The A block contained 40 trials, with a target step-1
amplitude of 10° or 20° and a step-2 amplitude representing 25% of
step-1. In the B block (60 trials), three amplitudes of target displacements
were used (7.5°, 15°, and 22.5°), and step-2 amplitude was 33% of
step-1. In the C block (40 trials), the amplitude of step-1 was either
12.5° or 25°, and the amplitude of the backward jump was 40% of
step-1. Note that the three step-2 amplitudes led us to choose specific step-1
amplitudes in accordance with the 2.5° spacing of LEDs along the
horizontal axis of our target board. Moreover, we progressively increased
step-2 amplitude in order to minimize the conscious perception of step-2 by
the subjects and to induce a larger adaptation. Indeed, an initial 40% step-2
would be consciously detected by subjects who may thus voluntarily compensate
for it. Note further that subjects, when asked at the end of experiments,
never reported an intrasaccadic step except the two non-naive subjects B and
C, who sensed the intrasaccadic step for the 40% amplitude. In each of the
three blocks, the amplitude and direction of step-1 occurred randomly and were
equally represented (10 repetitions). In addition, single-step target trials
were inserted within each block (12 trials in A and C blocks; four in the B
block) to increase the unpredictability of the whole sequence of target
presentation. For these trials, target-1 was not displaced but was instead
turned off during the saccade, and 1800 msec following its presentation, it
reappeared at the same position and became the starting point of the following
double-step trial. The entire adaptation session consisted of two repetitions
of the A block, two repetitions of the B block, and six repetitions of the C
block.
Test sessions
A test session was performed before the adaptation session (pre-test), to
measure the initial saccade performances of the subjects, and at different
times after the adaptation session (post-tests), to measure the after-effects
of saccade adaptation and persistence of memory. In all test sessions, the
fixation point was always at the center of the target board. After a random
delay (1200 msec, 1800 msec, or 2200 msec), the fixation point was
extinguished and simultaneously a peripheral target (target-1) was
illuminated. The peripheral target position varied randomly between trials,
either to the right or to the left, at ±10° or ±20°
eccentricity. In this condition, target-1 did not step during the saccadic
response but instead was turned off to prevent the occurrence of any
de-adaptation (i.e., recovery from adaptation) during the post-tests. Then,
1800 msec after target-1 presentation (i.e., 1–1.5 sec after its
offset), the next trial was initiated by the presentation of the fixation
point at the center of the board. Thus a test session was composed of four
types of trials repeated 10 times each.
Experimental schedule
The adaptation session was performed on day 0 (D0). On day 0, each subject also performed a pre-test (D0pre) and two post-tests. The first post-test (D0a) was performed after about half of the adaptation session (240 double-step trials: two A blocks, followed by two B blocks, followed by one C block). The second post-test (D0b) was recorded after the remaining 200 adaptation trials were performed (five C blocks). To avoid fatigue in subjects, breaks of 1–2 min, with eyes closed, were inserted between blocks in the adaptation session and between the test and adaptation sessions. These measures on day 0 allowed us to test the efficacy and the immediate after-effects of the adaptation induced by the double-step target protocol.
To test the retention of adaptation, a post-test was run at different times after D0: on days 1 (D1), 5 (D5), 11 (D11), and 19 (D19). Between these different post-tests, the subjects were free to perform their usual daily activities inside and outside the laboratory. Two subjects (subjects C and E) were again tested on day 90 and on day 105, respectively. Indeed, after completion of the data analysis, we noticed that these two subjects exhibited a saccade gain that was still significantly reduced at D19 with respect to D0pre. For subject E, saccade gain was even smaller at D19 than at D11. The delayed post-tests in these two subjects revealed for subject E a gain increase at D105, close to the value at D11, suggesting that some uncontrolled factor could be responsible for the gain drop observed at D19 in this subject. For subject C, the gain also increased at D90, even if it remained significantly smaller than the pre-test gain. Thus, for these two subjects, the gain measured at D90 or D105 was substituted for the gain at D19.
Protocol in control experiment
Subjects B and C performed, respectively, 21 and 32 wk after D0, a control
experiment with three other subjects (F, G, and H). The design of the control
experiment and the number of trials were exactly the same as in the main
experiment. The only difference is that during the adaptation session of the
control experiment, eye velocity was not monitored online and no backward
target step was induced during saccades. Instead, this second step from
target-1 to target-2 occurred only 1800 msec after target-1 onset. Target-2
position then became the fixation point of the following trial. Since the
change from target-1 to target-2 occurred long after the saccade toward
target-1 was completed (>1 sec), this control protocol should not elicit
any sensori-motor adaptation (Shafer et
al. 2000; Fujita et al.
2002; see Introduction) but was adequate for testing fatigue
effects since the elicited main and corrective saccades were the same as in
the main experiment. This session, called pseudo-adaptation, was composed of
the same successive blocks as in the adaptation session, with rest periods
between blocks. As in the main experiment, this pseudo-adaptation session was
preceded by a pre-test and followed by a first post-test after five blocks (A,
A, B, B, C) and by a second post-test after the last five C blocks, all
performed on day 0. A post-test was also performed on days 1 and 5 (D1 and D5,
respectively).
Eye movement recordings and data acquisition
Horizontal and vertical eye movements were recorded by video-oculography using an EyeLink system (SMI), at a frequency of 250 Hz with an accuracy of 0.1°. A new calibration was performed on each day immediately before recording sessions. During calibration, subjects were required to sequentially look at five targets located at 0°, -27.5°, and 27.5° positions on the horizontal axis and at ±12.5° along the vertical axis.
Left eye position was processed online by an electronic circuit (low pass
filtering, differentiation, and thresholding) to generate a signal when the
velocity exceeded a threshold of 
50°/sec. This signal was used by a
Data Wave computer program (Berthoud) to trigger the intrasaccadic target step
in the adaptation session or to extinguish the target in the test sessions. In
addition to target presentation, this computer program sampled eye position
data (sampling frequency = 500 Hz), displayed eye movements after each trial,
and stored the data on disk for off-line analysis.
Data analysis
The horizontal and vertical components of the movements of the two eyes were analyzed off-line. After filtering (70 Hz cut-off frequency, finite impulse response [FIR] filter, residual noise level <0.5°), the initial and final positions of all saccades were marked on the basis of a velocity threshold of 40°/sec. These two markings were checked by the experimenter and could be manually changed. Trials with a primary saccade latency <100 msec or >1000 msec or trials with a saccade that was not correctly detected online or that was contaminated with a blink were eliminated. For all subjects, excluded trials represented an amount of 8.3% ± 5.9%. For the data analysis, only the saccade in response to step-1, i.e., the primary saccade, was considered. Moreover, we computed the data of the cyclopean eye as the average of the data of the right eye and of the data of the left eye. This choice was dictated by the fact that there were significant differences in gain between rightward and leftward saccades in the pre-test when each eye was analyzed separately (i.e., a larger gain for rightward versus leftward saccades for the left eye and the opposite gain asymmetry for the right eye). To avoid these baseline, eye-dependent, directional asymmetries of gain, which would have interfered with our analysis on the effect of saccade direction in saccade adaptation, we averaged the data of the two eyes.
The parameters described in the following were computed for each primary saccade. Saccade amplitude was computed as the difference between the initial and final positions of the eye, and saccade duration was obtained as the corresponding time difference. Saccade gain was calculated as the ratio between saccade amplitude and target step-1 amplitude. For each subject, the percentage of gain change related to the adaptation protocol was calculated as follows: [(mean post-test gain (at D0a, D0b, D1, D5, D11, or D19) - mean pre-test gain (D0pre))/mean pre-test gain] * 100. Further, in the main experiment, the amount of retention was computed as follows: [percentage gain change (at D1, D5, D11, or D19)/percentage gain change at D0b] * 100. In a first part of the Results section, data of rightward and leftward saccades were pooled together. In a second part, rightward and leftward saccades were differentiated in order to test whether the adaptation equally affected both directions.
For the main experiment, the duration, peak velocity and latency of saccades recorded at D0 were compared between the pre-test (D0pre) and the two post-tests (D0a, D0b) by a repeated-measures ANOVA with the within-subjects test session factor (n = 5 subjects). Concerning saccade amplitude, the pre-test (D0pre) gain was compared to the post-test gain recorded on different days (D0a, D0b, D1, D5, D11, D19) by using a repeated-measures ANOVA with test session as the within-subjects factor. We then investigated if saccade gain in test sessions could differ according to the amplitude of the target displacement (10° and 20°) by a repeated-measures ANOVA with two factors (test session and target amplitude). Similarly, the difference in saccade gain according to the leftward versus rightward direction of saccades was tested by a repeated-measures ANOVA with two factors (test session and saccade direction). For the control experiment, the same comparisons between the pre-test and the post-tests as those described above were made, and the statistical analyses used the same repeated-measures ANOVAs as in the main experiment (n = 5 subjects). Post-hoc comparisons were performed by means of Fisher LSD tests. Statistical significance level was set at P < 0.05.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
1 E-mail pelisson{at}lyon.inserm.fr; fax 33-472-91-34-01.
|
REFERENCES |
|---|
|
TOPABSTRACTResultsEffect of the double-step...Effect of the double-step...DiscussionRetention of saccade gain...Effects of adaptation on...Materials and MethodsREFERENCES |
|---|
Abrams, R.A., Dobkin, R.S., and Helfrich, M.K. 1992. Adaptive modification of saccadic eye movements. J. Exp. Psychol. Hum. Percept. Perform. 18:922 -933.[CrossRef][Medline]
Alahyane, N. and Pélisson, D. 2004. Eye
position specificity of saccadic adaptation. IOVS
45:123
-130.
