Exp Brain Res DOI 10.1007/s00221-014-3996-8
Research Article
Implied motion perception from a still image in infancy Nobu Shirai · Tomoko Imura
Received: 12 September 2013 / Accepted: 16 May 2014 © Springer-Verlag Berlin Heidelberg 2014
Abstract Visual motion perception can arise from nondirectional visual stimuli, such as still images (implied motion, cf. Kourtzi, Trends Cogn Sci 8:47–49, 2004). We tested 5- to 8-month-old infants’ implied motion perception with two experiments using the forced-choice preferential looking method. Our results indicated that a still image of a person running toward either the left or right side significantly enhanced infants’ visual preference for a visual target that consistently appeared on the same side as the running direction (the run condition in Experiment 1). Such enhanced visual preference disappeared in response to an image of the same person standing and facing the left/right side (the stand condition in Experiment 1), an image of the running figure covered with a set of opaque rectangles (the block condition in Experiment 2) (Gervais et al. in Atten Percept Psychophys 72:1437–1443, 2010), and an image of the inverted running figure (the inversion condition in Experiment 3). These results suggest that only the figure that implied dynamic body motion shifted the infants’ visual preference to the same direction as the implied running action. These findings demonstrate that even infants as young as 5 to 8 months old are sensitive to the implied motion of static figures. Keywords Infant · Implied motion · Visual development · Visual perception
N. Shirai (*) Department of Psychology, Faculty of Humanities, Niigata University, 2‑8050, Ikarashi Nishi‑Ku, Niigata 950‑2181, Japan e-mail: [email protected]‑u.ac.jp T. Imura (*) Department of Information Systems, Niigata University of International and Information Studies, 3‑1‑1, Mizukino, Nishi‑ku, Niigata 950‑2292, Japan e-mail: [email protected]
Introduction Visual motion perception contributes to the execution of various adaptive actions, such as locomotion (e.g., Gibson 1950), postural control (e.g., Lee and Aronson 1974), interception or avoidance of moving objects (e.g., Savelsbergh et al. 1991; Schiff et al. 1962), and so on. Indeed, no animal has vision without the ability to detect visual motion (Nakayama 1985); thus, visual motion perception is one of the most fundamental and important visual functions in animals, including human beings. Although we usually detect and perceive visual motion from visual stimuli that contain directional information, we can also perceive a dynamic visual scene from stimuli that have particular form information without any directional information. For instance, the dynamic glass pattern (Ross et al. 2000), which involves rapid sequential presentation of global form patterns (i.e., “Glass patterns;” Glass 1969) composed of dot pairs (dipoles), creates the perception of global motion; perceived motion corresponds to the global orientation of the dipoles in the Glass pattern, even though the sequence has no directional information. We are also able to perceive directional motion from static figures. A still image of an object with streaks oriented in the direction of the represented motion of the object provides a compelling impression of dynamic movement of the object; blurs in the direction of the represented movement (motion smear; Burr 1980; Harrington et al. 1980) or lines in parallel with the motion direction attached to the tail of the object (motion lines; e.g., Kawabe and Miura 2006; Kawabe et al. 2007; Francis and Kim 1999) serve as cues for the perception of the object’s motion. Additionally, a more complex shape for the object itself can also be a strong cue for motion perception. When we see a still image of a person, an animal, or possibly an inanimate
13
object, we can extract a dynamic state of the subject from its postural figure in the still image (implied motion; e.g., Kawabe and Miura 2006; Kourtzi 2004; Winawer et al. 2008). It has been suggested that visual form cues that generate a directional impression aid in the perception of directional information from “ambiguous” visual motion signals (cf. Burr 2000; Burr and Ross 2002; Kourtzi et al. 2008). Motion perception is an important ability, and such ability seems to be mediated not only by directional motion but also by non-directional form visual processing. Although quite a few psychological and psychophysiological studies with human infants have shown that the ability to detect visual motion can be observed very early in life, there has been little evidence of significant motion perception ability from form information in early development. Previous studies indicated that significant directional sensitivity emerges within the first few months of life (e.g., Banton et al. 2001; Wattam-Bell 1991; Wattam-Bell 1996a, b), and even the sensitivity to relatively complex visual motion patterns such as radial optic flow can be observed by 3–4 months of age (Brosseau-Lachaine et al. 2008; Shirai et al. 2004a, b, 2006, 2008, 2009). On the other hand, emergence of motion perception from form information was empirically confirmed in much later developmental stages. Friedman and Stevenson (1975) tested development of the perception of implied motion in preschoolers (about 4 years of age), first- and sixth-grade elementary school children, and adults (college students). Visual stimuli were still pictures that schematically represented various human actions (e.g., moving arm(s) or leg(s), locomotion, bending the upper body) with motion lines, postural expression, multiple overdubbing of the figures in the sequence of the particular action (thus, this cue inevitably contained postural information), or none of these cues. Participants were asked to judge whether each picture seemed to be “moving” or “still.” The results indicated that adults and older elementary school children tended to categorize the figures with motion lines, postural expression, and overdubbing as “moving.” On the other hand, the preschoolers and first graders judged figures with a postural shape and overdubbing, but not others, as “moving.” Similar results with children aged 3–4 years were also reported by a later study (Carello et al. 1986). Thus, the ability to utilize form information (i.e., the postural state of a person) to perceive directionality develops until at least around 4 years of age. The basic abilities to detect both global form and motion patterns develop by 4–5 months of age (cf. Braddick and Atkinson 2007). Hence, infants or children younger than those participating in the previous studies (Carello et al. (1986); Friedman and Stevenson 1975) would potentially have the ability to process form–motion interaction such as implied motion perception. Indeed, several types of form–motion interactions, such as perceiving the shape of
13
Exp Brain Res
a moving object under slit-viewing (e.g., Parks 1965) conditions or perceiving a partially occluded moving object, were observed in infants aged 5 or more months (Imura and Shirai, submitted; Otsuka et al. 2009). These studies demonstrated infants have the ability to perceive form from motion. However, no studies have examined infants’ ability to perceive motion from form, such as implied motion perception. It is reasonable to assume that the ability to extract directional information from static form is also observed at similar ages. On this basis, we aimed to investigate whether young infants can perceive implied motion in the present study. We adopted modified versions of the visual stimulus, and experimental procedure of a recent psychophysical study with adult observers conducted by Gervais et al. (2010). They demonstrated that a still image of a person that has a postural shape representing a particular motor action, such as running or throwing, automatically shifts the observers’ visual attention toward the implied direction of the action. For instance, when a still image of a man running toward either the right or the left side was presented, the detection latency for a subsequent target that appears on the same (or the other) side of the implied direction was shortened (or delayed). They also found that a picture of the same man just standing and facing the lateral side (thus, the picture seemed not to imply dynamic action) did not elicit the attention shift. Their results indicate that implied dynamic body action, such as running and throwing, in a static picture promotes a visual attention shift toward the implied direction of the action. We examined whether similar behavior occurs in infant observers in order to investigate infants’ implied motion perception ability.
Experiment 1 Methods Participants The final sample included data from 15 5- to 6-monthold (mean age = 165.5 days, SD = ±19.8) and 15 7 to 8-month-old (mean age = 221.3 days, SD = ±15.3) infants. All participants were healthy full-term infants. An additional three infants participated in the experiment, but were excluded from the final sample because they could not complete all experimental trials due to crying. Apparatus All experiments were controlled by PsychoPy2 (Peirce 2009), software designed for psychological experiments, running on a personal computer (Apple, MacBook Pro
Exp Brain Res
2300/15 MD103 J/A). Visual stimuli were displayed on a 22-inch CRT display (Mitsubishi, RDF223H; refresh rate = 60 Hz, resolution = 1,024 × 768 pixels, full-color mode) in a dark experimental booth. Two speakers were positioned behind the CRT to attract infants’ attention to the CRT at the beginning of each trial by making beep sounds. Infants sat on their parents’ lap in front of the CRT. The viewing distance was approximately 40 cm. A TV monitor was placed outside the booth and connected to a CCD camera attached below the CRT so that an experimenter could observe the infants’ looking behavior on the monitor. Video outputs from the CCD camera were also forwarded to a video recorder, and all infants’ looking behaviors during experiments were recorded. Stimuli and procedures The present study was approved by the Ethical Committee for Psychological Research of Niigata University and was conducted according to principles established by the Helsinki declaration. Written informed consent was obtained from the parents of the infants. The infants were selected from the infant laboratory participant database at Niigata University (Niigata, Japan). Recruitment leaflets were distributed to the families of the infants via the Public Health Service Center and its branches in Niigata city. The database was composed of only infants whose families voluntarily contacted the infant laboratory. Each test trial began with the presentation of a colorful cartoon figure at the center of a white presentation field (width = 57.3 deg., height = 42.9 deg) accompanied by salient beep sounds. When an infant looked at the cartoon, an experimenter presented a cue figure at the center of the presentation field: a full-colored still image of a young male either running (run condition) or standing (stand condition) facing towards one of the lateral sides (see Fig. 1). Although both the running and standing cue figures were generated from photos of the same person at the same focal distance, the size of each figure was slightly different due to the posture of each; the width and the height were 8.4 deg × 15.0 deg for the running cue and 3.5 deg × 16.6 deg for the standing cue. It should be noted that Gervais et al. (2010) reported that a figure exhibiting throwing posture elicited an attention shift with a shorter latency than did a figure with running posture, although both postures elicited a significant attention shift. In the present study, however, we adopted the running, but not the throwing posture as a cue figure. It is plausible that young infants encounter people who are running more frequently than those who are throwing things. That is, we decided to use the posture that was potentially more familiar to young infants. After 600 ms from the onset of the presentation of the cue figure, two targets (black disks, diameter = 2.5 deg
Fig. 1 Examples of the cue figures used in the present study
each) simultaneously appeared on both the left and the right sides of the cue figure. The distance from the edge of each cue figure to the center of the presentation field was 17.2 deg. After the target presentation, the experimenter, who was naive to the stimulus identity and observed the infants’ looking behaviors on the TV monitor, judged in real time whether they looked first at the left or the right side of the presentation field by pressing one of two relevant keys on the personal computer (the forced-choice preferential looking technique; Teller 1979). The direction of the infant’s first gaze after the onset of the presentation of the two targets was used as the index for the judgments. If the experimenter judged that the direction of the infant’s gaze was consistent with the direction cued by the cue figure, the trial was counted as a consistent trial (for an overview of the procedure, see Fig. 2). There were two experimental conditions in Experiment 1: the run condition, which included the running cue figure, and the stand condition, which included the standing cue
13
Exp Brain Res
Fig. 2 Flow chart of the experimental procedure
figure. All infants took part in both conditions. We tested each infant with 20 trials under each condition, and the two conditions were presented separately. That is, trials under one of the two conditions were first repeated 20 times with a counterbalancing of the cue-figure direction (left/right × 10 trials) in random order, and then trials of the other condition were repeated in the same manner. Hence, each infant participated in a total of 40 trials. The order of the experimental conditions was counterbalanced across infants. We calculated a preference score for cued direction for each infant based on the experimenter’s judgments. The preference score was the ratio of the number of consistent trials to the number of all trials under each experimental condition. It would be expected that the running figure would shift the infants’ looking behavior to the implied running direction, and thus the preference score would be greater than chance (0.5) if infants were sensitive to visual information related to implied motion. We also calculated the inter-observer agreement for judgments of infant looking behaviors. Five experimental sessions (a total of 100 trials) were randomly chosen from each of the run and stand conditions, and a coder who was not the experimenter and who was naïve to the experimental conditions, aims of the experiment, positions of visual stimuli, and so on, judged infants’ looking behaviors in those trials based on offline movies. The observers agreed in 94 and 96 % of the run and stand trials, respectively. Results and discussion Infants’ preference scores under each experimental condition are displayed in Fig. 3a. We first conducted Bonferroni-corrected two-tailed one-sample t tests for the preference scores to test whether the preference scores were significantly different from chance (0.5). The preference scores were significantly higher than chance under the run condition for both
13
the 5 to 6- and the 7 to 8-month-old infants (t(14) = 5.78 p
Research Article
Implied motion perception from a still image in infancy Nobu Shirai · Tomoko Imura
Received: 12 September 2013 / Accepted: 16 May 2014 © Springer-Verlag Berlin Heidelberg 2014
Abstract Visual motion perception can arise from nondirectional visual stimuli, such as still images (implied motion, cf. Kourtzi, Trends Cogn Sci 8:47–49, 2004). We tested 5- to 8-month-old infants’ implied motion perception with two experiments using the forced-choice preferential looking method. Our results indicated that a still image of a person running toward either the left or right side significantly enhanced infants’ visual preference for a visual target that consistently appeared on the same side as the running direction (the run condition in Experiment 1). Such enhanced visual preference disappeared in response to an image of the same person standing and facing the left/right side (the stand condition in Experiment 1), an image of the running figure covered with a set of opaque rectangles (the block condition in Experiment 2) (Gervais et al. in Atten Percept Psychophys 72:1437–1443, 2010), and an image of the inverted running figure (the inversion condition in Experiment 3). These results suggest that only the figure that implied dynamic body motion shifted the infants’ visual preference to the same direction as the implied running action. These findings demonstrate that even infants as young as 5 to 8 months old are sensitive to the implied motion of static figures. Keywords Infant · Implied motion · Visual development · Visual perception
N. Shirai (*) Department of Psychology, Faculty of Humanities, Niigata University, 2‑8050, Ikarashi Nishi‑Ku, Niigata 950‑2181, Japan e-mail: [email protected]‑u.ac.jp T. Imura (*) Department of Information Systems, Niigata University of International and Information Studies, 3‑1‑1, Mizukino, Nishi‑ku, Niigata 950‑2292, Japan e-mail: [email protected]
Introduction Visual motion perception contributes to the execution of various adaptive actions, such as locomotion (e.g., Gibson 1950), postural control (e.g., Lee and Aronson 1974), interception or avoidance of moving objects (e.g., Savelsbergh et al. 1991; Schiff et al. 1962), and so on. Indeed, no animal has vision without the ability to detect visual motion (Nakayama 1985); thus, visual motion perception is one of the most fundamental and important visual functions in animals, including human beings. Although we usually detect and perceive visual motion from visual stimuli that contain directional information, we can also perceive a dynamic visual scene from stimuli that have particular form information without any directional information. For instance, the dynamic glass pattern (Ross et al. 2000), which involves rapid sequential presentation of global form patterns (i.e., “Glass patterns;” Glass 1969) composed of dot pairs (dipoles), creates the perception of global motion; perceived motion corresponds to the global orientation of the dipoles in the Glass pattern, even though the sequence has no directional information. We are also able to perceive directional motion from static figures. A still image of an object with streaks oriented in the direction of the represented motion of the object provides a compelling impression of dynamic movement of the object; blurs in the direction of the represented movement (motion smear; Burr 1980; Harrington et al. 1980) or lines in parallel with the motion direction attached to the tail of the object (motion lines; e.g., Kawabe and Miura 2006; Kawabe et al. 2007; Francis and Kim 1999) serve as cues for the perception of the object’s motion. Additionally, a more complex shape for the object itself can also be a strong cue for motion perception. When we see a still image of a person, an animal, or possibly an inanimate
13
object, we can extract a dynamic state of the subject from its postural figure in the still image (implied motion; e.g., Kawabe and Miura 2006; Kourtzi 2004; Winawer et al. 2008). It has been suggested that visual form cues that generate a directional impression aid in the perception of directional information from “ambiguous” visual motion signals (cf. Burr 2000; Burr and Ross 2002; Kourtzi et al. 2008). Motion perception is an important ability, and such ability seems to be mediated not only by directional motion but also by non-directional form visual processing. Although quite a few psychological and psychophysiological studies with human infants have shown that the ability to detect visual motion can be observed very early in life, there has been little evidence of significant motion perception ability from form information in early development. Previous studies indicated that significant directional sensitivity emerges within the first few months of life (e.g., Banton et al. 2001; Wattam-Bell 1991; Wattam-Bell 1996a, b), and even the sensitivity to relatively complex visual motion patterns such as radial optic flow can be observed by 3–4 months of age (Brosseau-Lachaine et al. 2008; Shirai et al. 2004a, b, 2006, 2008, 2009). On the other hand, emergence of motion perception from form information was empirically confirmed in much later developmental stages. Friedman and Stevenson (1975) tested development of the perception of implied motion in preschoolers (about 4 years of age), first- and sixth-grade elementary school children, and adults (college students). Visual stimuli were still pictures that schematically represented various human actions (e.g., moving arm(s) or leg(s), locomotion, bending the upper body) with motion lines, postural expression, multiple overdubbing of the figures in the sequence of the particular action (thus, this cue inevitably contained postural information), or none of these cues. Participants were asked to judge whether each picture seemed to be “moving” or “still.” The results indicated that adults and older elementary school children tended to categorize the figures with motion lines, postural expression, and overdubbing as “moving.” On the other hand, the preschoolers and first graders judged figures with a postural shape and overdubbing, but not others, as “moving.” Similar results with children aged 3–4 years were also reported by a later study (Carello et al. 1986). Thus, the ability to utilize form information (i.e., the postural state of a person) to perceive directionality develops until at least around 4 years of age. The basic abilities to detect both global form and motion patterns develop by 4–5 months of age (cf. Braddick and Atkinson 2007). Hence, infants or children younger than those participating in the previous studies (Carello et al. (1986); Friedman and Stevenson 1975) would potentially have the ability to process form–motion interaction such as implied motion perception. Indeed, several types of form–motion interactions, such as perceiving the shape of
13
Exp Brain Res
a moving object under slit-viewing (e.g., Parks 1965) conditions or perceiving a partially occluded moving object, were observed in infants aged 5 or more months (Imura and Shirai, submitted; Otsuka et al. 2009). These studies demonstrated infants have the ability to perceive form from motion. However, no studies have examined infants’ ability to perceive motion from form, such as implied motion perception. It is reasonable to assume that the ability to extract directional information from static form is also observed at similar ages. On this basis, we aimed to investigate whether young infants can perceive implied motion in the present study. We adopted modified versions of the visual stimulus, and experimental procedure of a recent psychophysical study with adult observers conducted by Gervais et al. (2010). They demonstrated that a still image of a person that has a postural shape representing a particular motor action, such as running or throwing, automatically shifts the observers’ visual attention toward the implied direction of the action. For instance, when a still image of a man running toward either the right or the left side was presented, the detection latency for a subsequent target that appears on the same (or the other) side of the implied direction was shortened (or delayed). They also found that a picture of the same man just standing and facing the lateral side (thus, the picture seemed not to imply dynamic action) did not elicit the attention shift. Their results indicate that implied dynamic body action, such as running and throwing, in a static picture promotes a visual attention shift toward the implied direction of the action. We examined whether similar behavior occurs in infant observers in order to investigate infants’ implied motion perception ability.
Experiment 1 Methods Participants The final sample included data from 15 5- to 6-monthold (mean age = 165.5 days, SD = ±19.8) and 15 7 to 8-month-old (mean age = 221.3 days, SD = ±15.3) infants. All participants were healthy full-term infants. An additional three infants participated in the experiment, but were excluded from the final sample because they could not complete all experimental trials due to crying. Apparatus All experiments were controlled by PsychoPy2 (Peirce 2009), software designed for psychological experiments, running on a personal computer (Apple, MacBook Pro
Exp Brain Res
2300/15 MD103 J/A). Visual stimuli were displayed on a 22-inch CRT display (Mitsubishi, RDF223H; refresh rate = 60 Hz, resolution = 1,024 × 768 pixels, full-color mode) in a dark experimental booth. Two speakers were positioned behind the CRT to attract infants’ attention to the CRT at the beginning of each trial by making beep sounds. Infants sat on their parents’ lap in front of the CRT. The viewing distance was approximately 40 cm. A TV monitor was placed outside the booth and connected to a CCD camera attached below the CRT so that an experimenter could observe the infants’ looking behavior on the monitor. Video outputs from the CCD camera were also forwarded to a video recorder, and all infants’ looking behaviors during experiments were recorded. Stimuli and procedures The present study was approved by the Ethical Committee for Psychological Research of Niigata University and was conducted according to principles established by the Helsinki declaration. Written informed consent was obtained from the parents of the infants. The infants were selected from the infant laboratory participant database at Niigata University (Niigata, Japan). Recruitment leaflets were distributed to the families of the infants via the Public Health Service Center and its branches in Niigata city. The database was composed of only infants whose families voluntarily contacted the infant laboratory. Each test trial began with the presentation of a colorful cartoon figure at the center of a white presentation field (width = 57.3 deg., height = 42.9 deg) accompanied by salient beep sounds. When an infant looked at the cartoon, an experimenter presented a cue figure at the center of the presentation field: a full-colored still image of a young male either running (run condition) or standing (stand condition) facing towards one of the lateral sides (see Fig. 1). Although both the running and standing cue figures were generated from photos of the same person at the same focal distance, the size of each figure was slightly different due to the posture of each; the width and the height were 8.4 deg × 15.0 deg for the running cue and 3.5 deg × 16.6 deg for the standing cue. It should be noted that Gervais et al. (2010) reported that a figure exhibiting throwing posture elicited an attention shift with a shorter latency than did a figure with running posture, although both postures elicited a significant attention shift. In the present study, however, we adopted the running, but not the throwing posture as a cue figure. It is plausible that young infants encounter people who are running more frequently than those who are throwing things. That is, we decided to use the posture that was potentially more familiar to young infants. After 600 ms from the onset of the presentation of the cue figure, two targets (black disks, diameter = 2.5 deg
Fig. 1 Examples of the cue figures used in the present study
each) simultaneously appeared on both the left and the right sides of the cue figure. The distance from the edge of each cue figure to the center of the presentation field was 17.2 deg. After the target presentation, the experimenter, who was naive to the stimulus identity and observed the infants’ looking behaviors on the TV monitor, judged in real time whether they looked first at the left or the right side of the presentation field by pressing one of two relevant keys on the personal computer (the forced-choice preferential looking technique; Teller 1979). The direction of the infant’s first gaze after the onset of the presentation of the two targets was used as the index for the judgments. If the experimenter judged that the direction of the infant’s gaze was consistent with the direction cued by the cue figure, the trial was counted as a consistent trial (for an overview of the procedure, see Fig. 2). There were two experimental conditions in Experiment 1: the run condition, which included the running cue figure, and the stand condition, which included the standing cue
13
Exp Brain Res
Fig. 2 Flow chart of the experimental procedure
figure. All infants took part in both conditions. We tested each infant with 20 trials under each condition, and the two conditions were presented separately. That is, trials under one of the two conditions were first repeated 20 times with a counterbalancing of the cue-figure direction (left/right × 10 trials) in random order, and then trials of the other condition were repeated in the same manner. Hence, each infant participated in a total of 40 trials. The order of the experimental conditions was counterbalanced across infants. We calculated a preference score for cued direction for each infant based on the experimenter’s judgments. The preference score was the ratio of the number of consistent trials to the number of all trials under each experimental condition. It would be expected that the running figure would shift the infants’ looking behavior to the implied running direction, and thus the preference score would be greater than chance (0.5) if infants were sensitive to visual information related to implied motion. We also calculated the inter-observer agreement for judgments of infant looking behaviors. Five experimental sessions (a total of 100 trials) were randomly chosen from each of the run and stand conditions, and a coder who was not the experimenter and who was naïve to the experimental conditions, aims of the experiment, positions of visual stimuli, and so on, judged infants’ looking behaviors in those trials based on offline movies. The observers agreed in 94 and 96 % of the run and stand trials, respectively. Results and discussion Infants’ preference scores under each experimental condition are displayed in Fig. 3a. We first conducted Bonferroni-corrected two-tailed one-sample t tests for the preference scores to test whether the preference scores were significantly different from chance (0.5). The preference scores were significantly higher than chance under the run condition for both
13
the 5 to 6- and the 7 to 8-month-old infants (t(14) = 5.78 p
