You are on page 1of 6

461

Central neural mechanisms for detecting second-order motion


Curtis L Baker Jr
Single-unit neurophysiology and human psychophysics have to this stimulus because of the linear summation of excita-
begun to reveal distinct neural mechanisms for processing tion and inhibition from its ON and OFF zones. However,
visual stimuli defined by differences in contrast or texture stimuli composed of fine-scale texture elements (carrier)
(second-order motion) rather than by luminance (first-order that have a large-scale variation (envelope) in contrast
motion). This processing begins in early visual cortical areas, (Figure lb) or spatial offset (Figure lc) have equal
with subsequent extrastriate specialization, and may provide a amounts of dark and light regions in each receptive field
basis for form-cue invariant analyses of image structure, such as zone, so the linear summation would give no net response.
figure-ground segregation and detection of illusory contours. Thus, these stimuli would be second-order, with respect to
the spatial scale of the neuron's receptive field. (However
Addresses stimuli constructed with large texture elements compared
McGill Vision Research Unit, Department of Ophthalmology, McGill to the size of the receptive field [4"] or with a net lumi-
University, 687 Pine Ave W, H4-14, Montreal, Quebec, Canada nance different from the background might still give
H3A 1 A 1; e-mail: curtis@astra.vision.mcgill.ca
first-order responses.) Although complex cells in Vl
Current Opinion in Neurobiology 1999, 9:461-466 behave nonlinearly, they nevertheless exhibit the same
kinds of luminance spatial selectivity as simple cells and
http://biomednet.com/elecref/0959438800900461
cannot explain response to second-order stimuli; therefore,
© Elsevier Science Ltd ISSN 0959-4388 an additional kind of nonlinear processing is necessary.
Abbreviations
A17 primary visual cortex of cat (see V1) This review will not only highlight some important results
A18 secondary visual cortex of cat (see V2) from the past two years, concentrating on the detection of
fMRI functional magnetic resonance imaging contrast-modulated pattern motion (such as in Figure lb),
MT/V5 extrastriate visual area specialized for motion
which is the best studied kind of second-order processing,
PET positron emission tomography
V1 primary (striate) visual cortex of primate (see A 17) but also include discussion of some pertinent studies using
V2 secondary (pre-striate) visual cortex of primate (see A 18) second-order texture stimuli lacking contrast modulation
VP upper hemifield counterpart of V3 (such as Figure lc).

Introduction Separate processing of first- and second-order


Natural images contain a rich mixture of surfaces and stimuli
boundaries of objects that are delineated by local differ- The detection of second-order stimuli may not necessarily
ences in luminance, contrast, texture, and motion. Much of require a separate processing mechanism from that used for
our understanding of visual perception has been achieved first-order stimuli. The simplest explanation for detection
using stimuli such as bars, edges, or sinewave gratings, of contrast modulation is that small nonlinearities early in
which are defined by vanat1ons in luminance. visual processing cause a slightly differing magnitude of
Neurophysiology has demonstrated that single neurons in response to dark and light regions of carrier texture, giving
the primary and secondary visual cortices can be remarkably a small net difference between high- and low-contrast
selective for local stimulus orientation, spatial frequency, regions that could be detected by subsequent linear filter-
and direction of motion, and human psychophysical experi- ing. Such an early nonlinearity, possibly occurring in the
ments have provided evidence for the independent early photoreceptors or artifactually in the stimulus apparatus,
detection of these luminance-defined (first-order) stimulus could plausibly .produce only a very weak signal (if it were
attributes. Recent advances in the experimental analysis of a strong signal, its consequences would preclude the body
the detection of borders defined by non-luminance (i.e. sec- of findings that led to the prevalent linear filtering theory).
ond-order, or non-Fourier) variations, such as differences of However, such a 'distortion product' could be a legitimate
contrast or texture [1,2], have accelerated our progress in concern in view of the very high contrast sensitivity of the
understanding the underlying mechanisms. visual system [5]. Artifactual distortions introduced by dis-
play monitors can be compensated with careful calibration,
The processing of luminance-defined (first-order) stimu- and a diffusing screen can be used as a spatial low-pass fil-
lus attributes is conventionally understood in terms of ter to assess the effectiveness of this correction [6,7""]. The
models in which selectiviry for stimulus properties such as more serious issue of early visual nonlinearity has been
spatial frequency or orientation is attributed to linear fil- investigated using a variety of psychophysical methods and
tering, with subsequent nonlinear gain control and stimuli [7"",8]- these studies concur that such early non-
thresholds affecting the amplitude of response [3"]. linear effects, probably arising in photoreceptors [9], are
Figure la shows an image of a sinewave grating stimulus small in magnitude and only become large enough to affect
upon which the receptive field of a Vl simple cell has been psychophysical data at higher temporal frequencies. A
superimposed. The Vl cell depicted would respond well related concern lies in inadvertent activation of first-order
462 Sensory systems

Figure 1

Examples of (a) first-order and (b-f) second·


order visual stimuli. (a) Conventional ( a) Luminance grating ( b) Contrast envelope (c) Illusory contour
luminance-defined sinewave grating. Light
and dark regions of the stimulus are aligned
with excitatory and inhibitory regions of the
superimposed schematic receptive field of a
V1 simple cell, such that linear spatial
summation would give a good response.
(b) One frame of a contrast-modulated
envelope stimulus, constructed with a
stationary high spatial frequency carrier
grating whose contrast is modulated by a (d) Texture-defined bar (e) Dynamic noise bar
moving low spatial frequency envelope.
Because the carrier spatial frequency is
relatively high, the light and dark image
regions cancel one another within excitatory
or inhibitory subzones of the simple receptive
field, giving no response from a linear, first-
order mechanism. (c) Abutting line gratings,
spatially offset where they meet, which give a
percept of an 'illusory contour' (used in
[24,25,29••]). A receptive field responding to
C urrent Opinion in Neurobiology
the vertical contour would represent another
kind of second-order response. (d) Bar·
shaped envelope enclosing a noise texture, envelope enclosing a region of dynamic stimulus, as in (b), but with carrier grating at
whose mean luminance is matched to the noise, against a background of static noise different orientation than envelope (used
backgrou nd (used in [29••]). (e) Bar-shaped (used in [23,26••,38]). (f) Contrast envelope in [1 o••]).

processing - particularly using second-order stimuli con- in both simple and complex cells of V 1/A 17 and, more
structed with relatively large, static texture elements. This commonly, V2/A18. Responses are found to illusory con-
problem can be avoided in neurophysiology by using a car- tour stimuli (Figure 1c) in monkey V2 [24,25],
rier texture of much finer grain than the measured linear noise-defined bars (Figure 1e) in monkey Vl [26""], con-
filter properties of the neuron [6,10""], and in psy- trast e nvelopes (Figure 1b) in cat A17 and A18
chophysics by e mploying dynamic carriers [4•]. [6,10 .. ,27,28"], and texture -defined bars (Figure 1d) in cat
A18 [29""]. In many of these studies, at least some neurons
Distinct mechanisms for processing first- and second-order respond as strongly to second-order as to luminance-
stimuli have been implicated in many psychophysical defined stimuli, making it very unlikely they are mediated
studies. Nishida eta/. [11""] report that adaptation to one by early nonlinearities.
kind of stimulus does not affect motion thresholds for the
other, and Schofield and Georgeson [1 2.. ] found that Construction of second-order stimuli from sinewave grat-
detection thresholds for each can be facilitated in an order- ing components has provided a powerful approach for
specific manner. We [13"] found that chromatically defined analyzing the neural signal processing of second-order
motion is directly detected only by second-order motion stimuli. Using contrast envelope stimuli (e.g. static, high
mechanisms. Several groups have reported that contrast spatial frequency sinewave grating carriers modulated by
envelope stimuli are best detected in a lower range of tem- drifting, low spatial frequency sinewave envelopes; see
poral frequencies than first-order stimuli [14,15",16"]; Figure 1b), our group has made a detailed analysis of neu-
other kinds of second-order stimuli show slower temporal rons in cat A17 and A18 [6,10"",27,28"], revealing neurons
dynamics [17- 19]. The two kinds of motion also exhibit that respond to both luminance gratings and second-order
differing degrees or kinds of vulnerability to interference envelopes. A very surprising result is that these neurons
from other added masking stimuli [20,2 1,22"]. are narrowly tuned for quite high values of carrier spatial
freq uency - not only much higher th an the spatial
These kinds of evidence, as well as recent neurophysiology response range for first-orde r gratings, but approaching the
(see below), indicate that first- and second-order stimuli are animal's acuity limit. The lower spatial resolution of visual
processed by separate, early mechanisms, before being neurons in cats compared to monkey makes it an advanta-
combined into a unitary motion representation. geous choice of species for these experime nts. T hese
neurons are also bandpass tuned to the spatial and tempo-
Responses to second-order stimuli in early ral frequency of envelope motion, with preferred
visual cortical areas frequencies somewhat lower than those for first-orde r
Neurons responding to second- as well as first-order stim- stimuli [28"]. Optimal envelope spatial frequencies are 8-
uli are present not only in extrastriate cortex [23], but also to 30-fold lower than optimal carrier frequencies, rather
Central neural mechanisms for detecting second-order motion Baker 463

Figure 2

Model of cortical neuron response, having


separate parallel signal-processing pathways
Linear filter
for first- and second-order stimuli.
Nonlinearities such as output thresholds and
gain controls [3•] are omitted for clarity. In the
top pathway, luminance-defined, first-order 1st order
stimuli are processed conventionally with a
linear spatiotemporal filter. The filter is
depicted as overlaid on a first-order stimulus,

~~
as in Figure 1a. The bottom pathway consists
of a serial 'filter-7rectify-7filter' cascade Response
Stimuli
consisting of early linear filtering subunits, a
nonlinearity (such as rectification), and a late
linear filter; the early and late filters are shown
superimposed on a contrast envelope
stimulus, to indicate the action of their
summation regions. The early filter controls
selectivity to the carrier texture, and the late
filter determines envelope selectivity. Note
that a first-order luminance grating could not
pass both the early and the late filters,
because of their very different spatial
selectivities. The neuron's response reflects a Current Opinion in Neurobiology
combination of both signaling pathways.

than being in a fixed ratio, as suggested by some psy- its orientation or direction of motion. It is the selectivity
chophysical studies of texture perception [30,31]. Optimal for properties of the envelope that is the essential second-
envelope temporal frequencies average about 4 Hz, con- order response provided by the later-stage filter.
sistent with human psychophysics [7••].
In experiments independently varying carrier and envelope
That the same neuron can show distinct optimal spatial orientations (Figure lt) [to••], envelope-responsive neurons
frequencies for luminance gratings, envelopes, and carriers showed the same preferred orientations for e nvelopes and for
of envelopes, is inconsistent with explanations based on luminance gratings; however, preferred carrier and envelope
early nonlinearities or other models having only a single orientations had no systematic relationship to one another.
kind of processing for both first- and second-order stimuli. The carrier orientation tuning suggests that neurons corre-
These and other findings can be understood in terms of a sponding to the early filter of the filter~rectify~filter
more elaborate filtering model having two signal-process- stream (Figure 2) are in the visual cortex.
ing pathways (Figure 2), one responding to first- and the
other to second-order stimuli [ l ,27], whose outputs are Some psychophysical studies have supported the neuro-
combined. One pathway responds conventionally to first- physiological findings of carrier spatial frequency selectivity
order stimuli, with a linear filter that determines spatial -masking of contrast envelope motion that is specific to
and temporal frequency selectivity to luminance gratings. carrier spatial frequency [22•] and a failure of apparent
The other pathway has an early linear filter that deter- motion when successive exposures are alternately defined by
mines carrier spatial frequency selectivity, t hen a first- and second-order patterns [7"•,36]. However, in second-
nonlinearity, such as full-wave rectification (giving like- order apparent motion of grating-like micropatterns [13•,37],
signed responses to both positive and negative signals), the carrier spatial frequency or orientation can be changed on
and a late filter that determines envelope spatial and tem- successive flashes, suggesting that some second-order mech-
poral frequency tuning. Such ' filter-Hectify~filter' anisms may have broadband early filtering.
models had previously been proposed in spatial vision for
textural figure-ground segregation [32-34] and for extrac- Extrastriate specialization for first- and
tion of contrast modulations contained in plaid stimuli second-order motion processing
[35]. In texture segregation models, the first fi lter picks out As first- and second-order information could be useful for dif-
the fine grain of a texture, and the second filter responds ferent visual functions, it would not be surprising if
to boundaries between texture regions. specialized extrastriate brain areas make differential use of
them. Neurons in primate extrastriate area MT (VS) and
Note that in the 'fi lter~rectify~fi lter ' processing scheme nearby areas show responses to dynamic noise bar stimuli
of Figure 2, the early-stage filters are doing a first-order (Figure le) [23,38], texture-defined stimuli [39], and contrast
operation in their response to the carrier; however, their envelopes [40], whereas inferior temporal cortex cells respond
response does not signal attributes of the e nvelope, such as to texture-defined stimuli [4 1]. It is difficult to conclude on
464 Sensory systems

the basis of these studies, however, whether extrastriate visu- envelope responsive neurons to carrier spatial frequency
al areas are specialized for first- or second-order processing. and orientation [10••,27] represents a lack of invariance that
Evidence for such specialization has come instead from neu- suggests some modification of the concept.
ropsychology and neuroimaging approaches.
In principle, there are some natural situations in which
Because only a fraction of cortical neurons have been found second-order information should be particularly useful,
to respond to second-order stimuli, a nonspecific brain such as disambiguating shadows, handling transparency
impairment might appear relatively more damaging to sec- [SO], or detecting moving, texture-covered objects whose
ond- than to first-order processing [42]. Vaina and constituent texture motion is irrelevant (such as pouncing
colleagues have reported the particularly striking finding of predators, fleeing prey, or approaching soccer balls) [51].
two patients, one having a relatively greater deficit for per- The filter-Hectify~filter processing scheme thought to
ception of second-order stimuli [43] and the other having a subserve second-order detection is potentially a very gen-
deficit specific for first-order stimuli ([44•]; see also [45]). eral and powerful way to extract image structure not
Anatomic MRI showed that these patients have brain revealed by conventional linear filtering, such as illusory
lesions in different extrastriate cortical regions [46••]. This contours and other kinds of texture borders. Extraction of
'double dissociation' lends further support to the notion second-order information may be an important element in
that the two kinds of stimuli are processed separately, and subsequent computation of higher-order image structure,
strongly suggests extrastriate specializations for them. such as plaids [35], terminators [52•], some forms of com-
plex motion [53,54•], and non-Cartesian analysis [Ss••].
Neuroimaging is a promising approach for exploring
extrastriate specialization. However, a potentially serious A potential problem with this processing scheme is the
problem is to ensure that a measured response to a second- very large number of combinations of filters needed to pro-
order stimulus is not merely attributable to first-order vide a complete basis set for representing higher-order
mechanisms responding to the carrier texture, rather than image structure. This problem may be less serious than it
to the envelope or texture boundary. While this question is seems as a result of constraints such as the limited range of
straightforward to disambiguate for a single, narrowly optimal carrier spatial frequencies and the broad band-
tuned neuron [6,10••], it is more challenging for the broad- widths of optimal carrier orientations [lo••,27,28•]. Also, in
band population response of a brain region. Using fMRI to view of the broadband nature of natural images, a truly
measure responses to a variety of first- and second-order complete representation may not be necessary.
motion stimuli in humans, Smith eta/. [4]-•] found that Vl
and most identified extrastriate visual areas respond to Conclusions
both; however, one area (identified as V3) responded more While the scheme of separate signal-processing pathways
to second- than to first-order motion, or to the carrier pat- for first- and second-order information is now well support-
tern presented alone, suggesting a degree of relative ed by human psychophysics, single-unit electrophysiology,
specialization. Van Oostende eta/. [48] report that fMRI and evidence for extrastriate specialization, there have not
responses to motion-defined stimuli selectively activate a been any reports of single neurons that respond exclusively
different extrastriate area. Using PET imaging of respons- to second-order stimuli. This discrepancy could be a simple
es to plaid stimuli, and subtracting the responses to the consequence of the protocols commonly used to search for
grating components, Wenderoth eta/. [49•] obtained a net visually responsive neurons, in which initial assessments are
response consistent with the contrast modulation compo- performed with simple luminance-defined stimuli.
nent of the plaid, which they localized to an extrastriate
visual region. It has been an implicit assumption throughout this review
that there are common kinds of mechanisms underlying
Function of second-order processing second-order motion, texture segregation, and perception
In natural vision, second-order information (much like of illusory contours [Ss••], though much of the research on
chromatic information) does not usually occur in isolation, these phenomena has proceeded somewhat independent-
but with correlated luminance information. Thus, it should ly. Whether this is so, or to what extent these different
nqt be surprising that all neurons found to be responsive to kinds of stimuli are detected by fundamentally different
second-order stimuli are also responsive to first-order stim- mechanisms, is an important contemporary research ques-
uli. A popular and perhaps related interpretation of this tion. Indeed, the present distinction of first- and
result is in terms of 'form-cue invariance' [23], the notion of second-order motion may well stand in need of substantial
a mechanism selective for attributes such as contour orien- revision, as some have already suggested [56], as new
tation or direction of motion, regardless of the nature by results and insights emerge.
which the contour is defined (e.g. by differences of lumi-
nance, contrast, texture, etc). This idea is supported for Acknowledgements
envelope orientation and direction of motion My research is supported by Canadian Medical Research Council (MA9685)
and National Science and Engineering Research Council (OGP0001978)
[lo••,26••,27,29••], and roughly, if not exactly, for envelope grants. I am grateful to Robert Hess. Fred Kingdom. Tim Ledgeway, and
spatial frequency [28•]; however, the selectivity of contrast Nick Scott-Samuel for their comments on the manuscript.
Central neural mechanisms for detecting second-order motion Baker 465

References and recommended reading luminance mechanisms and that chromatically defined stimuli should not be
Papers of particular interest, published within the annual period of review, defined as 'first-order'.
have been highlighted as: 14. Derrington AM, Badcock DR, Henning GB: Discriminating the
direction of second-order motion at short stimulus durations.
• of special interest
Vision Res 1993, 33:1785-1794.
•• of outstanding interest
15. Derrington A, Cox M: Temporal resolution of dichoptic and second
1. Chubb C, Sperling G: Drift-balanced random stimuli: a general
• order motion mechanisms. Vision Res 1998, 38:3531-3539.
basis for studying non-Fourier motion perception. J Opt Soc Am
The authors demonstrate that human motion perception has a lower temporal
1988, 5:1986-2007.
resolution for second-order contrast envelopes than for first-order luminance
2. Cavanagh P, Mather G: Motion: the long and short of it. Spatial gratings. They also show that both kinds of motion mechanism are able to com·
Vision 1989,4:103-129. bine signals binocularly, consistent with processing in early visual cortical areas.

3. Carandini M, Heeger OJ, Movshon JA: Linearity and gain control in 16. Smith AT, Ledgeway T: Sensitivity to second-order motion as a
• V1 simple cells. In Cerebral Cortex, val. 13. Cortical Models. Edited function of temporal frequency and eccentricity. Vision Res 1998,
by Jones EG, Ulinski PS. New York: Kluwer Academic/Plenum; 38:403-41 0.
1999:401-443. Using second-order stimuli constructed to avoid first·order artifacts (e.g. sta·
This book chapter summarizes a popular model of first-order processing in tic noise carriers with small texture elements or dynamic noise carriers), the
V1 simple cells, in which stimulus selectivity is largely determined by linear authors find slower temporal dynamics than for comparable first-order grat-
spatiotemporal filtering followed by a threshold; the filter's gain is deter- ings. This result supports the notion of separate processing of first- and sec·
mined by pooled responses of nearby receptive fields. A model of a possi- and-order stimuli.
ble biophysical implementation of the filtering and gain control is described.
17. Georgeson MA, Harris MG: The temporal range of motion sensing
4. Smith AT, Ledgeway T: Separate detection of moving luminance and motion perception. Vision Res 1990, 30:615-619.
and contrast modulations: fact or artifact? Vision Res 1997, 37:45-62.
18. Boulton JC, Baker CL Jr: Dependence on stimulus onset
These authors demonstrate that psychophysical detection of commonly used
asynchrony in apparent motion: evidence for two mechanisms.
second-order motion stimuli having static carrier patterns can be mediated via
Vision Res 1993, 33:2013-2019.
artifactual first-order information. They show that this problem can be avoided
by using dynamic noise carriers or by high-pass filtering a static carrier. 19. Lin LM, Wilson HR: Fourier and non-Fourier pattern discrimination
compared. Vision Res 1996, 36:1907-1918.
5. Taub E, Victor JD, Conte MM: Nonlinear preprocessing in short-
range motion. Vision Res 1997,37:1459-1477. 20. Lu ZL, Sperling G: Contrast gain control in first- and second-order
motion perception. J Opt Soc Am [A] 1996, 13:2305-2318.
6. Zhou YX, Baker CL Jr: Envelope-responsive neurons in areas 17
and 18 of cat. J Neurophysiol1994, 72:2134-2150. 21. Sex PJ, Baker CL Jr: The effects of distractor elements on direction
discrimination in random Gabor kinematograms. Vision Res 1997,
7. Scott-Samuel NE, Georgeson MA: Does early non-linearity account
37:1761-1767.
•• for second-order motion? Vision Res 1999, 39:2853-2865.
These authors use luminance nulling methods to measure psychophysically 22. Derrington AM, Ukkonen 01: Second-order motion discrimination
the signal that could come from early nonlinear detection of second-order • by feature-tracking. Vision Res 1999, 39:1465-1475.
stimuli. They find that this contribution is weak and significantly affects psy- A stimulus is constructed so that first· and second-order components move
chophysical detection thresholds only at temporal frequencies above about in opposite directions, and the effects of an added masking stimulus are
8 Hz. These findings imply that a separate processing mechanism exists for evaluated. The most effective mask for second-order motion is one having
second-order stimuli at low temporal frequencies. the same carrier spatial frequency, consistent with the early filtering of a fil-
ter-->rectify-->filter model (Figure 2) being narrowband.
8. Holliday IE, Anderson SJ: Different processes underlie the
detection of second-order motion at low and high temporal 23. Albright TO: Form-cue invariant motion processing in primate
frequencies. Proc Roy Soc Land [Bioi] 1994, 257:165-173. visual cortex. Science 1992, 255:1141-1143.
9. He S, Macleod 01: Contrast-modulation flicker: dynamics and 24. Peterhans E, von der Heydt R: Subjective contours - bridging the
spatial resolution of the light adaptation process. Vision Res 1998, gap between psychophysics and physiology. Trends Neurosci
38:985-1000. 1991,14:112-119.
10. Mareschal I, Baker CL Jr: Cortical processing of second-order 25. Grosof DH, Shapley RM, Hawken MJ: Macaque V1 neurons can
•• motion. Vis Neurosci 1999, 16:1-14. signal 'illusory' contours. Nature 1993, 365:550-552.
Recording from cat A 18 neurons, we found that contrast·envelope-respon-
sive neurons showed narrow bandpass tuning for carrier spatial frequencies 26. Chaudhuri A, Albright TO: Neuronal responses to edges defined by
8-30·fold higher than the cell's optimal envelope spatial frequencies. Cells •• luminance vs. temporal texture in macaque area V1. Vis Neurosci
were also selective for carrier orientation, with broad bandwidths, and no 1997, 14:949-962.
systematic relationship between preferred orientations of carrier and enve- The authors report that neurons in monkey V1 are responsive to dynamic noise
lope. Some cells preferred carriers orthogonal to their envelopes, suggest- bars on static noise backgrounds (Figure 1e). Many of these cells showed a
ing that they respond to illusory contour stimuli (Figure 1c). similar orientation selectivity for both first· and second-order stimuli, and some
were also similarly direction selective. This study is significant for demonstrating
11. Nishida S, Ledgeway T, Edwards M: Dual multiple-scale processing the common occurrence of such 'form-cue invariant' responses in primate V1.
•• for motion in the human visual system. Vision Res 1997,
37:2685-2698. 27. Zhou YX, Baker CL Jr: Spatial properties of envelope-responsive
These psychophysical experiments show spatial frequency selective eleva- cells in area 17 and 18 neurons of the cat. J Neurophysiol1996,
tion of first-order motion thresholds only by adapting to first·order stimuli, 75:1 038-1 050.
and of second-order motion thresholds only by adapting to second-order
stimuli; little or no cross-adaptation was found. These results support sepa- 28. Mareschall, Baker CL Jr: Temporal and spatial response to
rate early motion mechanisms for processing the two kinds of stimuli. second-order stimuli in cat area 18. J Neurophysio11998,
80:2811·2823.
12. Schofield AJ, Georgeson MA: Sensitivity to modulations of We recorded the response of cat A 18 neurons to contrast envelope stimuli
•• luminance and contrast in visual white noise: separate mechanisms and found that these neurons have bandpass tuning to envelope spatial and
with similar behaviour. Vision Res 1999, 39:2697-2716. temporal frequencies. Most cells prefer somewhat lower spatial and tempo-
Psychophysical thresholds for detection of static first-order (luminance grating) ral frequencies for envelopes than for luminance gratings. Slower temporal
and second-order (contrast modulation) patterns were specifically facilitated dynamics for envelope responses were also seen in latency, as measured by
by background levels of the same order, but were not facilitated by stimuli of the slope of temporal phase versus frequency plots.
different order. The results were best fit with a model involving summation of
signals from separate pathways for first· and second-order processing. 29. Leventhal AG, Wang Y, Schmolesky MT, Zhou Y: Neural correlates of
•• boundary perception. Vis Neurosci 1998, 15:1107-1118.
13. Baker CL Jr, Boulton JC, Mullen KT: A nonlinear chromatic motion Reports that neurons in cat A 1 7 and A 18 respond to texture-defined bars
mechanism. Vision Res 1998, 38:291-302. (Figure 1d) and stimuli similar to illusory contours (Figure 1c). The cells
Using random Gabor kinematograms constructed of isoluminant chromatic exhibited similar orientation and direction selectivity for both first- and sec-
Gabor micropatterns, with added luminance masking noise, we found robust ond-order stimuli (i.e. form-cue invariance). The demonstration that the
second- but not first-order motion. Isochromatic stimuli, however, supported same neurons can respond to a variety of second-order stimuli, including
both first- and second-order motion. This result suggests that processing of illusory contours, supports the idea of a common mechanism for detection
first-order motion with chromatic stimuli is attributable to cross-activation of of these different stimuli.
466 Sensory systems

30. Sutter A, Sperling G, Chubb C: Measuring the spatial frequency motion, and the other showing the reverse pattern. The principal sites of
selectivity of second-order texture mechanisms. Vision Res 1995, damage are, in both cases, extrastriate, but in complementary brain
35:915-924. regions. These results constitute a powerful 'double dissociation' of loss-
es of first- and second-order processing, and argue for separate process-
31. Kingdom FAA, Keeble DRT: On the mechanism for scale invariance ing of the two kinds of stimuli.
in orientation-defined textures. Vision Res 1999, 39:1477-1489.
47. Smith AT, Greenlee MW, Singh KD, Kraemer FM, Hennig J: The
32. Sutter A, Beck J, Graham N: Contrast and spatial variables in
•• processing of first- and second-order motion in human visual
texture segregation: testing a simple spatial-frequency channels
cortex assessed by functional magnetic resonance imaging
model. Percept Psychophys 1989, 46:312-332.
(fMRI). J Neurosci 1998, 18:3816-3830.
33. Malik J, Perona P: Preattentive texture discrimination with early fMRI was used to map several known visual areas in individual human brains,
vision mechanisms. J Opt Soc Am [A]1990, 7:923-932. whose responses were then measured to first-order and a variety of second-
order stimuli. All areas examined were activated by both kinds of stimulus,
34. Bergen JR, Landy MS: Computational modelling of visual texture but in varying degree, with areas V3 and VP more strongly activated by sec-
segregation. In Computational Models of Visual Processing. Edited by ond-order stimuli.
Landy MS, Movshon JA. Cambridge, MA: MIT Press; 1991 :253-272.
48. Van Oostende S, Sunaert S, Van Heeke P, Marchal G, Orban GA: The
35. Wilson HR, Ferrera VP, Yo C: A psychophysically motivated model kinetic occipital (KO) region in man: an fMRI study. Cereb Cortex
for two-dimensional motion perception. Vis Neurosci 1992, 9:79-97. 1997, 7:690-701.
36. Ledgeway T, Smith AT: Evidence for separate motion-detecting 49. Wenderoth P, Watson JD, Egan GF, Tochon-Danguy HJ, O'Keefe GJ:
mechanisms for first- and second-order motion in human vision. • Second order components of moving plaids activate extrastriate
Vision Res 1994, 34:2727-2740. cortex: a positron emission tomography study. Neuroimage 1999,
37. Werkhoven P, Sperling G, Chubb C: The dimensionality of 9:227-234.
texture-defined motion: a single channel theory. Vision Res Human brain imaging with PET was used to show brain regions respon-
1993, 33:463-485. sive to plaids composed of two added sinewave gratings; responses to
the component gratings alone were subtracted. The resultant activation
38. Geesaman BJ, Andersen RA: The analysis of complex motion was interpreted as representing the net contrast modulation of the plaid.
patterns by form/cue invariant MSTd neurons. J Neurosci 1996, The surprisingly circumscribed brain regions of activation were in extras-
16:4716-4732. triate visual cortex.
39. Olavarria JF, DeYoe EA, Knierim JJ, Fox JM, van Essen DC: Neural 50. Daugman JG, Downing CJ: Demodulation, predictive coding, and
responses to visual texture patterns in middle temporal area of spatial vision. J Optical Soc Am [Optics & Image Science]1995,
the macaque monkey. J Neurophysiol1992, 68:164-181. 12:641-660.
40. O'Keefe LP, Movshon JA: Processing of first- and second-order 51. Sun H, Frost BJ: Computation of different optical variables of
motion signals by neurons in area MT of the macaque monkey. looming objects in pigeon nucleus rotundus neurons. Nat
Vis Neurosci 1 998, 15:305-317. Neurosci 1998, 1 :296-303.
41. Sary G, Vogels R, Kovacs G, Orban GA: Responses of monkey 52. Loftier G, Orbach HS: Computing feature motion without feature
inferior temporal neurons to luminance-, motion-, and texture- detectors: a model for terminator motion without end-stopped
defined gratings. J Neurophysiol1995, 73:1341-1354. cells. Vision Res 1999, 39:859-871.
42. Plant GT, Nakayama K: The characteristics of residual motion Computer simulation was used to demonstrate that a model that combines
perception in the hemifield contralateral to lateral occipital signals from first- and second-order motion processing can correctly signal
lesions in humans. Brain 1993, 116:1337-1353. the motion of terminators of lines, which are thought to be computationally
important in higher-level image analysis.
43. Vaina LM, Cowey A: Impairment of the perception of second order
motion but not first order motion in a patient with unilateral focal 53. Ahlstrom V, Blake R, Ahlstrom U: Perception of biological motion.
brain damage. Proc Roy Soc Land [Biol]1996, 263:1225-1232. Perception 1997, 26:1539-1548.

44. Vaina LM, Makris N, Kennedy D, Cowey A: The selective impairment 54. Gurnsey R, Fleet D, Potechin C: Second-order motions contribute
of the perception of first-order motion by unilateral cortical brain • to vection. Vision Res 1998, 38:2801-2816.
damage. Vis Neurosci 1998, 15:333-348. Second- as well as first-order stimuli are shown to be effective in eliciting
Using a wide variety of motion stimuli and psychophysical tasks, the authors vection, a sensation of self-motion induced by large-field optic flow patterns.
examined a brain-damaged patient that has a perceptual deficit that is spe- In contrast, motion aftereffects were much more effectively driven by first-
cific for first-order motion, leaving second-order motion largely intact. This order stimuli. These results indicate one possible high-level functional role of
result is especially surprising because one would expect a non-specific second-order motion.
deficit of motion perception to have a greater impact on second- than on
first-order processing. 55. Wilson H: Non-Fourier cortical processes in texture, form, and
•• motion perception. In Cerebral Cortex, val. 13. Cortical Models.
45. Greenlee MW, Smith AT: Detection and discrimination of first- and Edited by Ulinski PS, Jones EG. New York: Kluwer Academic/Plenum;
second-order motion in patients with unilateral brain damage. 1999:445-477.
J Neurosci 1997, 17:804-818. This book chapter gives an overview of a variety of ideas about how fil-
ter-?rectify-?filter models can be powerfully used in the analysis of image
46. Vaina LM, Cowey A, Kennedy D: Perception of first- and second structure, such as . texture boundaries, illusory contours, curvature, 'non-
•• order motion: separable neurological mechanisms? Human Brain Cartesian' features, glass patterns, plaids and faces.
Mapp 1999, 7:67-77.
Anatomic MRI is used to compare the lesions of two brain-damaged 56. Lu ZL, Sperling G: The functional architecture ofhuman visual
patients, one having a perceptual deficit for second- but not first-order motion perception. Vision Res 1995, 35:2697-2722.

You might also like