Symposium
Chairs: Ann M. Skoczenski and Rick O. Gilmore
Discussant: Richard N. Aslin
The rebirth of infancy research a generation ago began with the study ofvisual development. Since that time, there has been remarkable progress inour understanding of basic sensory, perceptual, and cognitive aspects ofinfant vision. Despite this notable progress, however, many fundamentalquestions about how infant vision develops and what drives visualdevelopment remain largely unanswered. Moreover, researchers who focus onsensory and perceptual questions often have few opportunities to engage ina dialogue with colleagues who approach vision from a more cognitiveperspective.This symposium will bring together researchers whose approaches to visualdevelopment are widely varied but who share a common interest inunderstanding the nature of vision early in life and in revealing themechanisms which shape its development. Presenters will address questionsconcerning visual development at several different levels of analysis,ranging from the development of neural activity that supports motion andform processing, to eye movement measures of object representations, andan investigation of the spatiotemporal properties of infant visualattention and visual expectations. What unifies these diverse topics isthe shared goal of understanding early functional vision and itsdevelopment from an information processing perspective. The concludingdiscussion will center on a discussion of what we do and do not yet knowabout early visual development. As a result, in addition to offering newviews on several enduring questions in infant vision, this symposium willfocus on an agenda for future research in a vital research domain.
Details of individual items:
paper
Beyond the primary or striate visual cortex, the processing of forminformation for object recognition, and the processing of motion andspatial relationships for the control of action, appear to be segregatedbetween distinct pathways, named from primate anatomy as the 'ventral' and'dorsal' streams respectively. The function of these extra-striatepathways can best be assessed and compared by using stimuli which requireglobal integration of form information and motion informationrespectively. For this purpose we have devised new 'form coherence' and'motion coherence' displays which can be used for functional brainimaging, psychophysical testing of children and adults, and behaviouraltesting of infants.Functional MR brain imaging of adults shows that the brain areasresponding to form and motion coherence are beyond striate cortex and showvery little overlap, although the anatomical division is more complex thanthe classic ventral/dorsal division derived from primates. Inschoolchildren aged 4-7 years, psychophysical form coherence thresholdsare close to adult values and show little change with age over this range,while motion coherence thresholds show improvement with age but are stillwell above adult levels at age 7. These results suggest the extra-striatedorsal stream function is refined by a relatively slow process ofdevelopment in middle childhood, compared with that of the ventral stream.As well as differential developmental courses, there appears to bedifferential vulnerability in development of the two streams. We findthat motion coherence thresholds are poor relative to form coherence inchildren with hemiplegia from perinatal brain damage, Williams Syndrome,and autism, suggesting that development of the dorsal spatial actionstream is more readily disrupted than that of the ventral objectrecognition pathway.In infants, discriminations based on dorsal stream properties (directionof motion and disparity) emerge between 7-14 weeks, later than simplepattern discriminations. However, in this age range, preferential lookingtowards direction-defined targets is more robust, and shows lowercoherence thresholds, than preferential looking for a form-coherencetarget. Thus, although motion processing shows a postnatal delay in itsemergence, and a slow period of later development, following its initialemergence in infancy it is relatively effective compared to global formprocessing in controlling behavioural responses. This developmentalcourse will be discussed in terms of possible constraints placed by neuraltiming and interactions, and the infant's visual ecology.Supported by Medical Research Council programme grant G7908507
paper
Past research has suggested that the basic building blocks of adultspatial vision are in place, if not fully mature, during the first fewmonths of life. However, there is growing evidence that these buildingblocks, including contrast senstivity, grating acuity, and broadorientation discriminations, are not sufficient to support complex formprocessing in young infants. Here we describe two studies which suggestthat some important aspects of form perception remain strikingly immaturethroughout at least the first year of life. Both studies employed thevisual evoked potential (VEP) technique, which provides estimates ofsensory thresholds from brain responses to visual stimulation.First, we examined the development of relative position sensitivity,measured with Vernier acuity, in 57 healthy infants. Sensitivity to therelative position of visual pattern elements is critical fordiscriminating objects of different shapes and textures. Adults areexquisitely sensitive to relative spatial position, in fact their Vernieracuity is known as a hyperacuity, because it is superior to visualresolution (grating acuity). However, infants VEP Vernier acuity is worsethan their VEP grating acuity throughout the entire first year of life. Byone year of age, grating acuity is within a factor of two of adultsperformance, while Vernier acuity is still more than a factor of ten worsethan adults.Second, we examined infants processing of visual patterns containingcomplex orientation information. To accomplish this, we used asimultaneous visual masking paradigm, in which we measured VEP responsesto a grating of increasing luminance contrast presented alone and, withvisual masks that were one-dimensional spatiotemporal white noise patterns(similar to bar code patterns). Two mask conditions were used: orthogonaland parallel to the test grating. Adults had orientation-specific maskingof contrast thresholds: the orthogonal mask had an insignificant effect onthreshold while the parallel mask elevated threshold at least 6-fold.Three- four- and five-month-old infants showed no orientation-specificity:orthogonal and parallel masks produced the same factor of thresholdelevation in all three age groups. Orientation-specific masking emergedbetween eight and 11 months of age but was still quite immature at 11months. While several studies using serial presentations ofsingle-orientation patterns have shown orientation-tuned responses ininfants by 6-12 weeks postnatal, our data suggest that, for some stimulusconditions, interactions between mechanisms tuned to differentorientations are immature throughout most of the first postnatal year.Some theories suggest that adult Vernier hyperacuity is accomplished usingthe multiple orientation cues that appear in a Vernier offset stimulus,and our data further suggest that infants poor Vernier acuity may be dueto their inability to accurately process simultaneously-presented multipleorientation information. Because cortical processing is thought todominate both orientation discriminations and Vernier acuity, our findingswill be discussed in the context of visual cortical development data, andin particular, new data from infant monkey visual cortical neuronssuggesting that non-classical receptive field properties develop late,relative to classical receptive fields (Movshon et al., 1999).Supported by National Institutes of Health (EY-06579) and TheSmith-Kettlewell Eye Research Foundation
paper
Piaget (1954) described the neonate's visual experience as a series offleeting images, and an object's existence as dependent on directperceptual contact. Infants were proposed to develop objectrepresentations only after extensive experience viewing and manipulatingobjects. Piaget based these claims on observations of infant responses invarious reaching tasks.More recently, experiments have employed eye movements to probe infants'representations of moving objects that are occluded and again revealed.Bower, Broughton, and Moore (1971) showed infants an object moving behinda screen on its path. Two-month-olds were reported to disrupt trackingwhen the object emerged too soon but not when it changed form, and5-month-olds responded to changes in either variable. However, someinvestigators (Meicler & Gratch, 1980; Muller & Aslin, 1978) failed toreplicate these results. Muller and Aslin showed that object speed andocclusion time determined disruption, and Mareschal, Harris, and Plunkett(1997) suggested that inconsistencies across studies were due to differentobject speeds.A need exists, then, for systematic testing of tracking as an index ofinfants' object representations. Five 4-month-olds were presented a ballmoving laterally behind a screen, six cycles per trial, for eight trials(the Predictable-Motion group). The ball moved at 7.2 deg/s, wascompletely occluded for .73 s, and was visible in its entirety at eitherside of the occluder for 1.07 s. A corneal-reflection eye trackerobtained precise information about eye movements. It was hypothesizedthat if infants represented the existence and trajectory of the occludedball, they would anticipate its reemergence. Infants fixated the ball onthe majority of trials (79.5%). Anticipations (defined as eye movementstoward the side opposite where the ball became occluded, initiated atleast 100 ms before reemergence) comprised only 17.5% (SD 10.5) oftrials. Responses (eye movements toward the ball after reemergence), incontrast, comprised 58.4% (SD 9.5) of trials. This difference wassignificant, t(4) 5.15, p < .01. The infants did not, therefore,exhibit consistent anticipation despite repetitive motion, and pastreports of infants' visual expectations (Haith, 1993). Rather, mostfixations occurred after reemergence.To explore whether anticipations were chance occurrences, a second groupof four 4-month-olds viewed a moving ball with a random trajectory, suchthat side of reemergence was unpredictable (the Random-Motion group).Anticipations comprised 15.3% (SD 15.0) of trials, which did not differrelative to the Predictable-Motion group, t(7) .25, ns. Finally, athird group of four 4-month-olds viewed four trials with no screen,followed by four trials with predictable motion behind a screen (thePrior-Experience group). Anticipations comprised 30.0% (SD 13.4) of thelast four trials, a marginally significant difference relative to the lastfour trials of the Predictable-Motion group, t(7) 2.08, p .076.Because many anticipations may have been random, the eye movementsobserved in this study do not appear to reflect a robust representation ofthe ball's continued existence and subsequent reemergence. When givenexperience with a predictable, unoccluded trajectory, however, infantsretained this information and engaged in more anticipations. Thesefindings provide evidence that 4-month-olds' perceptual experience may beclose to what Piaget had originally proposed: a world in which objects donot have enduring representations after becoming occluded.
paper
The function of eye movements is to bring interesting stimulus events ontothe fovea for detailed visual, attentional, and perceptual processing.Studies have found that a predictable alternating spatial pattern of eventoccurrence facilitates 3-month-old infants' eye movements by increasingtheir anticipatory visual orienting to the locations of events beforetheir onset (e.g., Haith, Hazan, & Goodman,1988). These findingsdemonstrate that infants form expectations for the spatial parameter ofstimulus events. Stimulus events, besides occurring in space, also occurin time. Two experiments were conducted to determine whether infants'expectations and, consequently, their eye movements are sensitive to thetemporal environment in which events occur.In Experiment 1,13-week-old infants viewed a spatially alternating(left-right) sequence of pictures whose temporal onset predictablyalternated between 800 and 1200 msec. Infants made more anticipatory eyemovements in the temporally alternating condition than a temporally randomcondition (800 or 1200 msec), indicating that expectations for eventoccurrence is sensitive to the temporal environment. An analysis ofwhether infants formed expectations for the specific temporal onset ofstimulus events failed to reveal evidence that infants delayed theiranticipations longer during a 1200 msec ISI than during a 800 msec ISI, aswould be expected if they had formed expectations for specific temporalonsets. Results indicated, however, that infants used the previous trial'stemporal onset as a cue for timing their anticipatory orienting to thesubsequent trial, suggesting that infants might encode the average flow(timing) rate of event information.To test this possibility, in Experiment 2, infants viewed a spatiallyalternating event sequence in which picture durations and interstimulusintervals varied randomly among three different temporal values with theconstraint that the information flow rate for each two event segment(i.e., left picture-ISI-right picture-ISI) equalled a fixed temporal valuewhich was long (4200 msec), intermediate (3400 msec), or short (2600msec). Later in the event sequence, the two-event information flow ratewas switched to a new value: Infants who first received the long averagerate were switched to the shorter, intermediate rate and infants who firstreceived the short average rate were switched to the longer, intermediaterate.If infants form expectations for temporal rate at which events flow, thenswitching from the long rate value to the shorter one should result in areduction in the number of anticipatory eye movements because the eventswould more often precede the expectation for their later temporal onset.Infants who were switched from the short information flow rate to thelonger one should have the opposite effect--an increase in the number ofanticipations due to an expectation for an earlier temporal onset of anevent. Results were consistent with these predictions. These resultsreveal that infants' visual orienting, information-processing, andexpectations are based on the encoding of the average rate at whichinformation flows in relation to the temporal stream of events, ratherthan based on an individual, discrete event in temporal isolation.