Who is Crossing Where?: Infants’ Discrimination of Figures and
Grounds in Events
Tilbe Göksun1,2, Kathy Hirsh-Pasek1, Roberta Michnick Golinkoff3, Mutsumi Imai4, Haruka
Konishi3, and Hiroyuki Okada5
1Temple University
2University of Pennsylvania
3University of Delaware
4Keio University
5Tamagawa University
Abstract
To learn relational terms such as verbs and prepositions, children must first dissect and process
dynamic event components. This paper investigates the way in which 8- to 14-month-old Englishreared
infants notice the event components, figure (i.e., the moving entity) and ground (i.e.,
stationary setting), in both dynamic (Experiment 1) and static representations of events
(Experiment 2) for categorical ground distinctions expressed in Japanese, but not in English. We
then compare both 14- and 19-month-old English- and Japanese-reared infants’ processing of
grounds to understand how language learning interacts with the conceptualization of these
constructs (Experiment 3). Results suggest that 1) infants distinguish between figures and grounds
in events; 2) they do so differently for static vs. dynamic displays; 3) early in the second year,
children from diverse language environments form nonnative - perhaps universal - event
categories; and 4) these event categories shift over time as children have more exposure to their
native tongue.
Keywords
Event perception; cross-language comparison; prelinguistic constructs; figure and ground
Verbs and prepositions express relationships between the figures and grounds that unfold in
events. Thus, when we talk of a skater (a figure) who glides across the ice (the ground), the
specific action of glides entails a figure and a ground and is distinguishable from other
potential interactions like tripping, hopping or slipping. The learning of relational terms like
verbs is central to language acquisition because verbs form the fulcrum around which a
sentence is constructed. Learning these words, however, is difficult because infants must not
only parse events into components like figures and grounds but also “package” these
components in ways that are aligned with their native tongue. By way of example, English
© 2011 Elsevier B.V. All rights reserved.
. Address for correspondence: Tilbe Göksun, Department of Neurology, University of Pennsylvania, 3400 Spruce Street 3 Gates,
Philadelphia, PA 19104, tilbe@mail.med.penn.edu.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our
customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of
the resulting proof before it is published in its final citable form. Please note that during the production process errors may be
discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
NIH Public Access
Author Manuscript
Cognition. Author manuscript; available in PMC 2012 November 1.
Published in final edited form as:
Cognition. 2011 November ; 121(2): 176–195. doi:10.1016/j.cognition.2011.07.002.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
rarely conflates ground information within the verb (consider the verbs swim and fly) but socalled
“ground verbs” in Japanese routinely encode the spatial configuration of the ground
being traversed (e.g. wataru ‘go across’ implies that someone crosses a flat barrier dividing
two points such as a road or a railroad track) (Muehleisen & Imai, 1997).
This research is at the intersection of event processing and language development. We ask
how children process basic components of events at a time when they are at the cusp of
word learning, and also when most children have amassed a native vocabulary of 50 or more
words. When do infants demonstrate an ability to parse events into foundational components
like figures (i.e., prominent agent undergoing the motion) and grounds (i.e., a reference
point or a stationary setting) and how does exposure to their native language influence
toddlers’ interpretation of these event components?
Infants Process Components of Events
During the first year, infants detect an object’s motion (Haith, 1980), discriminate changes
in patterns of motion (e.g., Bogartz, Shinskey, & Schilling, 2000), and use motion to parse
actions in events (Baldwin, Baird, Saylor, & Clark, 2001; Hespos, Saylor, & Grossman,
2009; Spelke, Born, & Chu, 1983; Sharon & Wynn, 1998; Wynn, 1996). Once infants
perceive the actions within events, they must also detect those aspects of events that are
related to linguistic expressions (Clark, 2003).
Cognitive linguists (Jackendoff, 1983; Lakoff, 1987; Langacker, 1987; Talmy, 1985, 2000)
and developmental psychologists (e.g., Mandler, 1992, 2004) have long proposed that a set
of prelinguistic constructs is foundational to learning relational terms. A dynamic event is
composed of semantic components such as path (the trajectory of the motion), manner (how
an action is performed), source (the starting point), and goal (the endpoint) (Jackendoff,
1983; Talmy, 1985) that are labeled across the worlds’ languages. Other foundational
constructs refer to the spatial relations between objects such as containment (putting things
in a container) and support (putting things on a surface) (Choi & Bowerman, 1991).
Research on the way that infants and adults process events for language is relatively recent
(Casasola & Cohen, 2002; Choi, 2006; Göksun, Hirsh-Pasek, & Golinkoff, 2010; Golinkoff,
1981; Golinkoff & Hirsh-Pasek, 2008; Lakusta, Wagner, O’Hearn, & Landau, 2007; Malt &
Wolff, 2010; Mandler, 1992, 2004, 2010; Parish-Morris, Pruden, Ma, Hirsh-Pasek, &
Golinkoff, 2010; Pruden, Hirsh-Pasek, Maguire, & Meyer, 2004; Pruden, Göksun,
Roseberry, Hirsh-Pasek, & Golinkoff, in press; Pulverman, Golinkoff, Hirsh-Pasek, &
Buresh, 2008; Shipley & Zacks, 2008; Wagner & Carey, 2005; Wagner & Lakusta, 2009).
Thus far, the spatial and event components studied share a set of common characteristics
(Göksun et al., 2010). First, they are all perceptually available (e.g. one can detect the
manner of gliding by witnessing the event above) (Mandler, 2004). Second, because these
components are represented in the world’s languages, they seem to be universal (e.g., all
languages seem to mark paths and manners in events; Jackendoff, 1983; Talmy, 1985).
Third, and importantly, even though they are expressed universally, there are languageparticular
ways of encoding these semantic components in verbs and prepositions. For
example, some languages tend to express path information with prepositions and manner
information within verbs (e.g., run out in English), whereas others codify the same path of
motion within the verb by using an optional manner in the adverb (e.g., saler corriendo ‘exit
running’ in Spanish).
Linguistic descriptions of the semantic components of sentences are good starting points for
uncovering what infants might know about dynamic events. The burgeoning literature in this
area suggests that preverbal children notice and categorize spatial and event components
such as path, manner, source, goal, containment, and support that are codified in verbs and
Göksun et al. Page 2
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
prepositions (Casasola & Cohen, 2002; Choi, 2006; Lakusta et al., 2007; Pruden et al., 2004;
Pulverman et al., 2008). Pulverman and her colleagues, for example, habituated English- and
Spanish-reared infants (14- to 17- months) to an animated starfish performing both a path
and a manner (e.g., a starfish twisting as it moves over a ball). Using a within subject
design, infants saw 4 kinds of events: control (i.e., same event as habituation), path change
(the starfish twisting under the ball), manner change (the starfish spinning over the ball),
and path and manner changes (starfish flapping past the ball). Both English- and Spanishreared
infants increased their attention to path changes and manner changes relative to the
control event, suggesting that they isolated these components within events (Pulverman et
al., 2008).
These findings are intriguing given the differential “manner bias” in English. By way of
example, English has been estimated to have many more manner verbs compared to Spanish
or Turkish (Slobin, 2005). When describing short event clips (e.g., a boy crawling up a low
hill or a girl jumping into a pool), English speakers produced 18 times more manner verbs
than path verbs (Naigles, Eisenberg, Kako, Highter, & McGraw, 1998). Results from a
recent study suggest that before 3 years of age, English-, Spanish-, and Japanese-speaking
children assume that a verb labels the path represented in the event (Maguire et al., 2010).
By age 3 and beyond, they manifest language-specific patterns of verb construal such that
English-speaking children are more likely to map a novel verb to the manner of the motion,
compared to Spanish- and Japanese-speaking children. Perhaps infants initially and
universally extract the same information from the events that they witness and later, once
language is processed, attend differentially to the semantic components of events that are
highlighted in their language.
Some support for this notion also comes from the study of containment and support
relations. Containment is defined as the relationship that occurs when something is fully or
partially surrounded by a container and it is captured by the English word in. Support refers
to the contact of an object on a surface and is illustrated by the English word on. Using both
looking times and reaching behavior as dependent variables, Baillargeon and her colleagues
demonstrated that infants discriminate between events in ways that demonstrate an
understanding of containment and support (e.g., Aguiar & Baillargeon, 1999; Baillargeon,
Needham, & DeVos, 1992; Hespos & Baillargeon, 2001a, 2001b, 2006, 2008; Wang,
Baillargeon, & Paterson, 2005).
If infants’ event perception starts from a common base, we might expect that prelinguistic
infants will be sensitive to these spatial distinctions – even if they are not lexicalized in their
native language. This concept-to-language hypothesis proposes that event categories like
path and manner or containment and support would be acquired before language has its
influence (e.g., Göksun et al., 2010; Hespos & Spelke, 2007). As language meets dynamic
events, it may dampen attention to some event components and highlight others. And
indeed, this is what the literature suggests. The finding that English-reared babies perceive
containment and support events that are expressed in Korean lends credence to this
perspective (Choi and Bowerman, 1991).
In a now classic study, Choi and Bowerman (1991) noted that Korean does not have
English-equivalent words for in and on. Instead, it encodes containment and support
depending on the degree of fit relation between two objects. For example, putting a ring on a
finger and putting a book in a cover are described with the verb kkita, signifying the tightfitting
relationship between two objects. The contrasting verb nehta connotes loose-fit
relations (i.e., put in, around or together loosely), such as putting a book on a table or putting
a fruit into a bowl (Bowerman & Choi, 2001; Choi, 2006).
Göksun et al. Page 3
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Hespos and Spelke (2004) asked whether English-reared 5-month-olds’ might distinguish
between these “Korean” tight- and loose-fit events in both containment and support relations
even though they are not expressed in their ambient language. After familiarization (e.g., an
object fits tightly into a container), the infants were presented with both a familiar relation
(tight-fit) and a novel relation (loose-fit). Infants looked longer at the novel relation,
suggesting a keen sensitivity to tight- and loose-fit relations. By the second year after birth,
responses from English- and Korean-speaking children diverged in how they processed kkita
(‘tight-fit in’) versus nehta (‘loose-fit in’) spatial relations. Even though English-speaking
children at the ages of 29 and 36 months decreased in sensitivity to the difference between
tight- and loose-fit containment events, Korean-speaking children maintained those
distinctions. Hence, language-specific aspects of these spatial categories influenced
children’s nonlinguistic sensitivity at least by 29 months of age. In addition, Englishspeaking
29-month-old children with more words in their vocabularies relative to their
peers, or the ability to produce the word in, were less likely to perceive the difference in the
Korean degree-of-fit relation as compared to low vocabulary children or to those who did
not yet produce the word in (Choi, 2006). Thus, exposure to native language environment,
coupled with children’s knowledge of the specific prepositions that encode these relations,
negatively correlate with the detection of non-native semantic distinctions.
This concept-to-language perspective can be contrasted with an extreme view that we might
term the language-to-concepts hypothesis, suggesting that children are prompted to parse
nonlinguistic events only as they learn language. This hypothesis favors Whorfian linguistic
relativity (Whorf, 1956), which proposes that language itself influences the way people
think. That is, language is a “tool” that enables children to find components in the events
they witness; learning a language might help in constructing new concepts (e.g., Bowerman,
2007; Bowerman & Levinson, 2001; Choi & Bowerman, 1991). A recent paper by Gentner
and Bowerman (2009) offers a middle ground approach proposing that some spatial
categories might exist prelinguistically, with others that are less salient and are represented
more rarely across languages demanding linguistic experience to be learned (see also
Gentner, 1982).
Taken together, the literature suggests that prelinguistic infants notice and conceptualize
spatial and event components in ways that are conducive to learning all of the languages in
the world regardless of their ambient language. This view parallels the research in infant
discrimination of all of the sounds of language before they home in on the particular
contrasts used in their native language (e.g., Eimas, Miller, & Jusczyk, 1987; Kuhl et al.,
1997; Kuhl, 2004; Werker & Tees, 1984). That is, infants might discriminate and attend to a
broad palette of event constructs that will later be refined and “semantically organized” with
respect to the native language (for reviews see Göksun et al., 2010; Hespos & Spelke, 2007).
To date, however, only a small set of event components relevant to learning relational terms
has been examined. This work expands that literature by probing three questions: 1) How do
English-reared infants process figures and grounds in dynamic and static representations of
events?; 2) Are infants from different language environments similarly sensitive to aspects
of events that are codified in languages around the world?; and 3) Does sensitivity to aspects
of events change as infants are exposed to their ambient language?
Figure and Ground
The relationship between a static figure and a static ground was initially studied by Gestalt
psychologists (e.g., Koffka 1935; Wever, 1927) in terms of segmenting figures from
grounds. Since then, the perception of figure and ground has been investigated in the
literature on adults’ event processing (e.g., Kimchi & Peterson, 2008; Peterson & Gibson,
Göksun et al. Page 4
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
1994). There is also research on infants’ perception of figure and ground relations (e.g.,
Johnson & Aslin, 1998; Johnson & Mason, 2002; Kauffman-Hayoz, Kauffman, & Stucki,
1986). However, more studies are needed to assess infants’ ability to differentiate between
dynamic figures on various grounds and in real life settings as a prerequisite for learning
relational terms.
Why is it important to study infants’ processing of dynamic figures with respect to grounds?
The discrimination of humans in dynamic events might be fundamental and associated with
the concepts of agency and animacy that are constructed in the first two years of life (Eimas,
1994; Mandler, 1992, 2004; for a review see Rakison & Poulin-Dubois, 2001). The
interpretation of other people as agents (Johnson, 2000), for example, is related to
understanding the role people play in causal events (e.g., Golinkoff, 1975, 1981; Golinkoff
& Kerr, 1978; Oakes, 1994; Poulin-Dubois & Schultz, 1990); the intentionality behind a
person’s movement (e.g., Phillips, Wellman, & Spelke, 2002; Woodward, 1998, 2003) and
the means agents use to attain goals (Sodian, Schöppner, & Metz, 2004).
The ground in a dynamic event is a reference point in the form of an object or the stationary
setting of the scene. Ground information is central for the linguistic encoding of motion
events. In English, for example, the prepositions over, into, through, and across specify both
a path that the figure follows and the spatial properties of the ground object. Hence, ‘into’
not only refers to a path that the figure moves along, but also indicates that the ground object
is some kind of enclosure (Talmy, 2000). When the ground is the setting where an action
takes place, different relational terms implicitly encode different grounds. For example,
‘across’ implies a relatively stable surface that can be traversed, while ‘along’ implies a
more or less horizontal principal axis (Jackendoff, 1992; Landau & Jackendoff, 1993).
Intriguingly, in some languages such as Korean or Japanese, ground information for
stationary setting is specified within the verbs (Choi-Jonin & Sarda, 2007; Muehleisen &
Imai, 1997). Japanese, for example, classifies motion path verbs into two categories:
directional-path and ground-path verbs. Directional path (DP) verbs define the direction of
motion relative to a starting point or goal (e.g., hairu ‘enter’, iku ‘go’, kaeru ‘return’, kuru
‘come’), and do not restrict the ground on which the motion occurs (Muehleisen & Imai,
1997). However, ground-path (GP) verbs such as wataru ‘go across’, koeru ‘go over’, and
nukeru ‘pass through,’ incorporate properties of the ground along with the direction of
motion (Beavers, 2008; Muehleisen & Imai, 1997; Tsujimura, 2006). The spatial geometry
of the ground is the key to assigning the correct verb to a motion event. For example, in the
sentence Jun wa kawa/michi o watatta ‘Jun crossed the river/street,’ the meaning of wataru
‘go across a flat barrier dividing two points’ implies that there is both a starting point and a
goal, and that the ground is a flat extended surface such as a street (see Figure 1). The verb
wataru could not be used to describe a person moving across grounds such as a field or a
tennis court, grounds that have no clear borders or barriers that demarcate the two sides of
the plane. Even though the tennis court is separated into sections, there is no clear barrier
between its sections or around it. To describe a crossing action on these grounds, a more
generic verb tooru ‘go across a continuous plane’ is used. Thus, even the same action that
takes place against different kinds of grounds would demand that a different verb be used.
Compared to DP verbs, these GP verbs are very specific in regards to the ground that they
encode. However, both animate and inanimate figures can be used (Muehleisen & Imai,
1997). In the current studies, we use only animate figures (i.e., humans) to show how
crossing actions take place on different grounds. The use of all animate figures heightens the
distinctions between the different kinds of crossing events without introducing another
variable.
Göksun et al. Page 5
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
The encoding of grounds in some Japanese verbs, but not in English verbs, offers an
opportunity to study another potential case in which event information is perceptually
accessible, universally expressed in languages, and differentially labeled across languages.
That is, as in the cases of containment and support or path and manner, the study of figures
and grounds allows us to ask whether Japanese and English infants, prior to learning much
language, make similar semantic figure and ground “cuts” when viewing dynamic events
and whether exposure to language influences which ground distinctions are dampened or
heightened when children learn their native tongue.
Using a nonlinguistic preferential looking paradigm (Golinkoff et al., 1987; Hirsh-Pasek &
Golinkoff, 1996), three experiments explore how English- and Japanese-reared infants
discriminate figures and grounds in events and how native language exposure influences
infants’ attention to these event components. In this paradigm, infants were first familiarized
with single events and then shown one novel and one familiar event simultaneously. If
infants can discriminate between figures and grounds in events, they should prefer to watch
the novel figure or ground at test. Four specific questions were addressed: First, can Englishreared
infants detect figures in nonlinguistic dynamic events (Experiment 1)? We predict
that infants will notice changes in the figures in dynamic events. Second, will English-reared
infants show sensitivity to the detailed distinctions of grounds encoded only in Japanese
(Experiment 1)? We predict that infants will be able to discriminate grounds with respect to
the spatial geometry that is represented differentially within verbs in Japanese even though
they are not coded in English. Third, is the detection of figure and ground different when
infants are processing dynamic events versus watching static scenes (Experiment 2)? We
hypothesize that infants will differentiate between figures and grounds earlier when both
components are static. Finally, how do English- and Japanese-reared toddlers diverge in
their discrimination of grounds as they learn language (Experiment 3)? We expect that only
Japanese-reared toddlers who are exposed to a language that uses ground distinctions
(wataru ‘flat barriers dividing two points’ vs. tooru ‘continuous plane’) will continue to
notice the distinction between these classes of grounds.
Experiment 1: Discrimination of Figures and Grounds in Dynamic Events
Experiment 1a: Can 8- and 11-Month-Olds Discriminate Figures and
Grounds in Dynamic Events?
English-reared infants’ discrimination of figures and grounds in a crossing event was
examined to ask whether infants notice language-relevant components in events even when
their ambient language does not encode them. The Japanese ground-path verb wataru ‘go
across a flat barrier dividing two points’ can be used with grounds such as a railroad track, a
road, a bridge, and a street that extends in a line and with particular starting and ending
points, but not with a tennis court or a grassy field that extend in a plane. Instead, the verb
tooru ‘go across a continuous plane’ is used to describe a figure as it moves across a tennis
court and a field. English, uses the verb ‘to cross’ appropriately with all six grounds. Thus,
the Japanese verb wataru ‘a flat barrier dividing two points’ is more specific and restricted
in meaning in comparison to the English verb ‘cross.’
Infants were familiarized with a dynamic scene in which a figure (e.g., a man) crossed a
ground (e.g., a road). At test, they were presented with the same event in a split-screen,
either with a change in the figure (e.g., a woman vs. a man crossing a road) or a change in
the ground (e.g., a man crossing a railroad track vs. crossing a road). Importantly, ground
comparisons were manipulated such that infants were presented with some grounds coded
by the same verb in Japanese (e.g. a railroad track and a road) and with some examples that
would be coded with two different verbs in Japanese: wataru ‘a flat barrier dividing two
Göksun et al. Page 6
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
points’ as in the road and tooru ‘go across a continuous plane’ as in the tennis court. That is,
infants were either presented with grounds that both extended in a line and were bounded (a
railroad track vs. a road) or one that extended in a line and was bounded, and another that
extended in a plane and was unbounded (a railroad track vs. a grassy field). Even though
there are other grounds encoded by the verb wataru ‘go across’ (e.g., an ocean or a river,
because these bodies of water also separate two points), for the current studies, we selected
only a perceptually salient subset of grounds that all extend in a line (a road, a street, a
bridge, and a railroad track).
Eight- and 11-month-old infants were tested in either a figure or a ground discrimination
study. These age groups were chosen because previous findings suggest that infants
discriminate various event components prior to producing their first words between 7 and 12
months of age (Lakusta et al., 2007; Pulverman et al., 2008). We hypothesized that 1) none
of the infants in the figure or ground discrimination study would have an a priori preference
before familiarization to a specific figure or ground; 2) infants should look longer to the
novel figure or novel ground in the test trial, if they notice the change from familiarization
trials to test trial; and 3) parallel to the degree-of-fit distinction made in Korean, Englishreared
infants would also differentiate grounds better when the comparison was between two
categories in Japanese (wataru ‘go across’ vs. tooru ‘go through’) as opposed to within the
same Japanese category. That is, infants might consider the road vs. railroad track
comparison more similar than the road vs. tennis court comparison.
Method
Participants
Fifty 8-month-old (M= 8.01, SD= .76, 24 males) and 45 11-month-old (M=11.04, SD= .82,
25 males) English-reared infants participated in this experiment. Infants were randomly
assigned to either the figure discrimination (26 8-month-olds and 21 11-month-olds) or the
ground discrimination study (24 8-month-olds and 24 11-month-olds). All infants were
monolingual and full-term at birth. Infants were predominantly Caucasian and of middleclass
families from two Northeastern cities in the United States. Infants who had been tested
at 8-months-old were not tested again at 11 months. An additional 26 infants across two age
groups were excluded from data analyses because they were bilingual (n = 7), premature (n
= 1), the data were lost due to experimental error (n = 1), low attention to video clips (n =
10), having a side bias (n =3, see below in the coding section for the criterion), or fussingout
during the experiment (n = 4).
Stimuli
The stimuli consisted of televised displays of four people (a woman, a man, a 6-year-old girl
and a 6-year-old boy) crossing one of the six grounds (railroad track, road, narrow street,
bridge, tennis court, and grassy field) from left to right. In a 175 × 125 pixel image of the
scene, children had an average height of 28 pixels and adults had an average height of 40
pixels. In each event, the figure crossed all the way across the ground two times, after one
crossing was completed, as the clip repeated again from the beginning of the event. The pace
of walking was controlled across the event clips. No linguistic audio accompanied these
dynamic events. Stimuli were videotaped outdoors. All figures and grounds are presented in
Figure 1a and 1b.
In the figure discrimination study, different conditions displayed one of the three
comparisons of figures: adult-adult, adult-child, and child-child. The figures were presented
on different grounds in different conditions. For example, an infant might have seen the
Göksun et al. Page 7
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
comparison of a woman and a man either both on a road or both on a tennis court. The use
of human figures enabled us to keep the clips roughly comparable.
In the ground discrimination study, two conditions emerged based on the encoding in
Japanese: within-category comparisons (wataru ‘a flat barrier dividing two points’ i.e., a
railroad track, a road, a narrow street, and a bridge) and across-category comparisons (e.g.,
railroad track and tennis court or road and grass). The “wataru” grounds signal clear
boundaries between the starting point and the goal point and extend in a line. A grassy field
and a tennis court are not proper grounds for the verb wataru ‘a flat barrier dividing two
points’; they are instead encoded by the verb tooru ‘a continuous plane’ in Japanese since
they possess no barriers to divide them from their surroundings.
Procedure
Infants were tested using the nonlinguistic Preferential Looking Paradigm (Golinkoff et al.,
1987; Hirsh-Pasek & Golinkoff, 1996; Pruden et al., 2004) where children are seated on
their parents’ laps approximately 2.5 feet from the front of a 44-inch television screen (34 ×
28 inches). So as not to influence the child’s direction of looking, parents were instructed to
keep their eyes closed during the study.
Two cameras were placed at equal distance from the sides of the television (15 inches). The
video camera on the right ran the movie for the study; the video camera on the left recorded
the child’s eye gaze for offline coding (for a similar set up, see Maguire et al., 2010). This
enabled a clear view of eye gaze to the left or right side of the screen. Once the experimenter
started the movie, he or she left the room to avoid influencing the infant’s attention. Each
movie lasted for 132 seconds. Upon completion of the study, infants received a small gift of
either a t-shirt or a toy.
The stimulus movie contained four main phases: introduction, salience, familiarization, and
test trials (for sample conditions see Figure 2). The trials were separated by an attention-
getter.
Introduction Phase—An animated character appeared first on one side of the screen and
then on the other to ensure that infants were familiarized to clips playing on both sides of the
screen. Each presentation was 6 seconds long.
Salience Phase—Infants first saw what was to later become the test trial. This was used
to determine whether there was any a priori preference for either clip. The salience phase
contained a 12-second clip of two events on a split-screen.
Familiarization Phase—Infants watched four 12-second clips of exactly the same
stimulus on the full screen that involved a figure crossing a ground (e.g., woman crossing a
railroad).
Test Phase—Infants watched the test events simultaneously on the split-screen for 12
seconds. In the figure discrimination study, infants were presented with the comparison
between same figure/same ground (a woman crossing a railroad) vs. novel figure/same
ground (a man crossing a railroad). In the ground discrimination study, the test trials
compared the same figure/same ground (a woman crossing a railroad) clip seen during
familiarization with the same figure/novel ground (a woman crossing a road) video clip. For
within-category comparisons, the grounds were all from the “wataru” category, such as a
railroad track vs. a road. For across-category conditions, the comparison used one ground
from the “wataru” category and one ground that could not be described by the Japanese
“wataru” (e.g., a road vs. a tennis court, or a railroad vs. a field).
Göksun et al. Page 8
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Attention-Getter—A 3-second smiling baby face accompanied by the children’s song
“Oh, Susanna” was used to separate the trials in each phase of the experiment. The attentiongetter
had two purposes: to renew infants’ interest in the movie and to reorient the infants’
looking to the center of the screen before they had to choose between pairs of stimuli in the
split-screen trials. The side of the novel figure and the novel ground was counterbalanced in
both salience and test trials. No linguistic stimuli or audio of any type accompanied the
clips. Infants’ looking times were recorded for later coding.
Coding
The dependent variable was total looking time towards each event. During attention-getters
and familiarization phases, infants’ looking to the center of the screen was measured. For the
introduction, salience, and test phases, attention to the left or right side of the screen was
coded using a button-press box (see McDonough et al., 2003).
Coders were always blind to condition to ensure that they did not know the target side in the
movie. Each infant was coded twice to obtain intra-individual reliability. Coders were
trained to consistently meet a standard of 99% reliability for both intra- and inter-judge
coding. The intra-rater reliability is .998 (SD = .004) for all participants; 41% of all
videotapes were coded by a second person for inter-coder reliability (r = .991, SD = .010).
Attention was calculated by taking an infant’s total visual fixation time during all phases and
dividing this number by the length of the entire movie. If infants looked at the movie less
than 50% of the time (i.e. “low attention”), the data were removed from the sample. Ten
infants were excluded due to this criterion. Side bias was calculated by dividing infants’
total looking to the right side of the screen by their total looking time to both right and left
sides of the screen. In this calculation, only split-screen phases were included. If the
calculation was greater than .80 to one side, this was taken as an indication of a side bias,
and the infant’s data would be excluded from analyses. Three infants were excluded due to
this criterion.
Results
Percentage of looking time towards each event in the split-screen was calculated for salience
and test trials. We report results separately for the figure and ground studies. As no
significant gender differences emerged for either study, we did not consider gender as a
separate factor for further analyses.
Figure discrimination
A repeated-measures ANOVA with age (8- and 11-month-olds) as the between-subject
variable and trials (salience and test) as the within-subject variable yielded a trial by age
group interaction, F (1, 45) = 7.14, p= .01, η2= .14. No main effects of trial or age group
were found. That is, in the salience trials infants did not have any significant a priori
preference for the event clips at any age: 8-month-olds: t (25) = 1.58, p = .13, 11-montholds:
t (20) = .43, p = .66. However, as seen in Figure 3 (top graph), only 11-month-old
infants looked significantly longer to the novel figure compared to the same figure in test
trials, t (20) = 4.53, p = .01 (two-tailed), which was also above chance level, t (20) = 4.20, p
= .001.
Infants in both age groups were equally attentive during the entire movie (71% and 74%,
respectively for 8- and 11-month-olds). Additionally, looking times during the
familiarization phase were examined to determine whether there was an age difference in
infants’ attention to the events. A repeated measures ANOVA with age (8- and 11-monthGöksun
et al. Page 9
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
olds) as the between-subject variable and 4 familiarization trials as the within-subject
variables yielded only a main effect of familiarization trial, F (1, 135) = 6.29, p= .01, η2= .
12, but no main effect of either age or an age by trial interaction. Finally, no significant
difference was found for different comparisons of figures (adult-adult, adult-child, or child-
child).
Ground discrimination
A repeated-measures ANOVA with age (8- and 11-month-olds) and ground condition
(within- and across-category) as the between-subject variables and trials (salience and test)
as the within-subject variable demonstrated no main effects of trial, age group, or ground
condition, nor any interactions among them, Fs < .1.14, ps >.23. Infants did not have any a
priori preference for either event at either age: 8-month-olds: t (23) = .92, p = .37, 11month-olds:
t (23) = .48, p = .64, and infants at these age groups had similar percentages of
looking times to novel and familiar grounds in the test trials in both within- and acrosscategory
comparisons (see Figure 3, bottom graph). No significant difference was found for
different comparisons of grounds in each condition (within- and across-categories; Figure
4).
We again did not find any significant difference in terms of infants’ attention during the
entire movie (74% and 76%, respectively for 8- and 11-month-olds). Looking time analyses
during the familiarization trials yielded no main effects of age or familiarization trial or an
interaction between them. In both age groups, although there was a decline in looking times
across the familiarization trials, it was not significant.
Discussion
The results from Experiment 1a suggest that by 11 months of age, English-reared infants
only noticed the change of figures in these events. That is, infants discriminate figure
changes earlier than ground changes in dynamic events. This is consistent with previous
studies on other conceptual precursors such as containment-support and source-goal, which
demonstrate that infants notice some components in events earlier than others. For example,
infants distinguish the goal of an action before the source of an action (Lakusta et al., 2007)
and containment events before support events (e.g., Casasola & Cohen, 2002).
Why do infants process figures earlier than grounds in dynamic displays? Figures are also
expressed earlier than other conceptual distinctions in children’s early utterance
combinations (e.g., Clark, 1979; Grace & Suci, 1981; Tomasello & Merriman, 1995). Bock
and Warren (1985) argue that the conceptual accessibility of what is related to a hierarchy of
grammatical relations and the mental representations of these most accessible concepts are
learned earlier. Research with toddlers has also demonstrated that visual attention was
focused on the agents in a dynamic scene (Robertson & Suci, 1980). Hence, it is not
surprising that in our study infants processed the human figures earlier than the grounds in
the scenes.
On a related point, the perceptual saliency of event components might be related to differing
developmental trajectories. Perhaps a moving, animate figure commands infants’ attention
more so than a stationary ground. Clark (2009) suggests that children’s early conceptual
categories are influenced by perceptual Gestalts. Thus, a figure is against a ground, and the
moving figure object would be more salient than any one part of the scene. From very early
on, infants are sensitive to motion and differentiate biological from nonbiological movement
(Bertenthal, 1993). In addition, infants perceive the unity of a center-occluded object when it
moves in front of a textured background surface, but they do not perceive object unity when
it lacks a background texture (Johnson & Aslin, 1996, 1998; see also Kellman & Spelke,
Göksun et al. Page 10
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
1983). By the end of the first year of life, infants represent animate beings as having goaldirected
movements, intentions, and as agents of causal events (e.g., Golinkoff & Kerr,
1978; Rakison & Poulin-Dubois, 2001; Woodward, 1998). The present study adds to this
literature, suggesting that infants detect people as figures on various grounds in dynamic
events.
Eleven-month-olds discriminated between all figure comparisons (adult-child, adult-adult,
and child-child), suggesting that they detect information other than a person’s height as a
sign of figure change. Why were 8-month-olds not successful in recognizing the change of
the person in these dynamic events? One possible explanation is that people were shown
laterally, perhaps making it harder for younger infants to differentiate between them. Even
though research suggested that 7-month-olds were more sensitive to actions than faces
(Bahrick & Newell, 2008), in dynamic scenes where people’s actions were similar (i.e.,
crossing different grounds), facial features might be a useful and salient cue for
discriminating between people.
Experiment 1b: Can 14-Month-Olds Discriminate Between Grounds in Dynamic Events?
Experiment 1a suggested that 11-month-old infants noticed the change of a human actor in
dynamic scenes on different grounds. In the ground discrimination study, however, we
found that 11-month-olds did not have a significant preference for novel or familiar grounds
at test. To examine the possibility of a developmental change in ground discrimination, we
recruited an older age group of infants: 14-month-olds. We predicted that 1) none of the
infants would have an a priori preference for specific grounds; 2) infants should look longer
to the novel ground in the test trial, if they distinguish the ground change between the
familiarization trials and the test trial; and 3) English-reared infants at this age would also
differentiate grounds better when the comparison was between two categories in Japanese
(wataru ‘go across’ vs. tooru ‘go through’). That is, these infants would notice ground
changes better in across-category comparisons.
Participants
Twenty-four 14-month-old (M= 14.22, SD= .92, 12 males) English-reared infants
participated. All infants were monolingual and full-term at birth. Infants were predominantly
White and belonged to middle-class homes in two Northeastern towns in the United States.
An additional 6 infants were excluded from data analyses due to experimenter error (n = 1),
infant low attention to video clips (n = 1), or fussing-out during the experiment (n =4).
Stimuli, Procedure, and Coding
All exactly the same as in the ground discrimination study in Experiment 1a.
Results and Discussion
Percentage of looking time towards each event in the split-screen was calculated for salience
and test trials. When no significant gender differences emerged, gender was not considered
as a separate factor for further analyses.
A repeated measures ANOVA with ground condition (within- and across-Japanese category)
as the between-subject variable and trials (salience and test) as the within-subject variable
showed only a main effect of ground condition, F (1, 22) = 4.55, p = .04, η2= .17. No
significant preference to either event was obtained at salience, t (23) = 1.17, p = .19.
However, infants looked significantly longer to the novel ground in the test trial, t (23) =
2.20, p = .03. As shown in Figure 5, infants looked significantly longer to the novel ground
but only in the across-category comparison (e.g., railroad vs. tennis court) at above chance
Göksun et al. Page 11
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
levels, t (12) = 4.48, p = .001. They did not show the same pattern in the within-category
comparison, p = .87. No significant difference was found for different comparisons of
grounds within each condition.
Lastly, looking times during the familiarization phase were examined to determine whether
infants attended to the events and whether there was a difference in attention between
within- and across-category comparisons. A repeated-measures ANOVA with condition
(within- vs. across-category) as the between-subject variable and four familiarization trials
as within-subject variables yielded only a main effect of familiarization trial, F(3, 66)= 7.59,
p= .01, ηp
2 = .27. No main effect of condition or an interaction between familiarization trial
and condition was found. This finding suggests that infants in both conditions were attentive
during the familiarization phase and infants’ looking times declined during familiarization
across the trials.
Results from experiment 1b demonstrate that English-reared infants differentiated between
grounds at 14 months of age. More importantly, when infants noticed ground changes, they
did so for distinctions not encoded in English (i.e., the categorical difference of wataru ‘go
across’ in Japanese). To what extent is this finding merely a product of the perceptual
differences between these stimuli? If only perceptual differences mattered, then any
variation in grounds should have led to looking time differences – even for comparisons of
the grounds in the within-category conditions. This was not the case. Instead, these findings
suggest that the geometry of grounds – despite perceptual differences - are better described
as falling into distinct categories, namely, wataru and tooru categories.
Our findings corroborate the growing body of research on event perception. Infants parse
events into their components and attend to distinctions in events even when these are not
codified in their native language. Results from the ground discrimination study suggest that
infants are sensitive to a categorical distinction between grounds made in Japanese but not in
English. In particular, the grounds coded by the verb wataru ‘go across’ share certain
geometric features: They are bounded and extend in a line with specific starting and ending
points. In contrast, other grounds do not have these features. English-reared infants attend to
these common features between wataru grounds, and notice the changes in grounds only
when the ground changes from one Japanese category to another. This is reminiscent of
English-reared 5-month-old infants’ sensitivity to the degree-of-fit relation encoded only in
Korean (Hespos & Spelke, 2004).
Experiment 1c: Can 14-Month-Olds Discriminate Between Grounds in Dynamic Events on
Grayscale?
The findings from Experiments 1a and 1b demonstrated that infants notice changes in
figures at 11 months and grounds at 14 months in dynamic motion events. In ground
discrimination, infants only noticed the novel ground when one ground was from the wataru
category and the other was either a tennis court or a grassy field, grounds not encoded with
the verb wataru ‘go across.’ Infants, for instance, differentiated a tennis court from a
railroad track – a Japanese cross-category comparison - better than they did a railroad track
vs. a road – a within-category Japanese comparison.
There is, however, a potential perceptual confound in the stimuli given the nature of the
videos used to test this distinction. As shown in Figure 1, in addition to the nature of the
geometry of the grounds, a tennis court and a grassy field have a salient color. One might
argue that the homogenous stretches of green or red of the grassy field and the tennis court,
respectively, allowed infants to categorize these two grounds together. To rule out this
possibility and to secure our interpretation, we conducted a control experiment in which
infants were presented exactly the same movies in grayscale. If color of the ground were a
Göksun et al. Page 12
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
strong perceptual cue for infants’ discrimination of grounds, we would not expect to obtain
ground category effects using a black and white screen. Although removing the color will
make the scene more artificial, findings from this control study would eliminate any
possibility that infants look longer to grounds because of surface perceptual features.
We hypothesized 1) that there would be no difference between events in the salience trial, 2)
if infants perceptually discriminate grounds according to the Japanese ground distinctions,
removing color from the scene should not change the results. For example, infants would
still find the comparison of road vs. grassy field (a cross-category comparison) easier to
detect than the comparison of road vs. railroad track (a within-category comparison). Thus,
we predicted that there would be no difference in ground differentiation between the color
screen (Experiment 1b) and the black-white screen used here.
Method
Participants
Twenty-four 14-month-old (M= 14.01, SD= .79) English-reared infants participated,
balanced for gender. All infants were monolingual and full-term at birth. Infants were
predominantly White and from middle-class homes in two Northeastern towns. An
additional 5 infants were excluded from data analyses due to bilingual exposure (n = 1) and
having low attention (n = 4).
Stimuli
The stimuli were exactly the same as with the ground discrimination study in Experiment 1a,
except that the movies were presented in grayscale. Again, there were two conditions based
on the Japanese GP distinctions: within- category (wataru ‘go across’, e.g., a railroad track
and a road) and across-category (e.g., a railroad track and a tennis court or a road and a
grassy field) comparisons.
Procedure and Coding
The procedure and coding were the same as the ground discrimination study in the
Experiment 1a. Twenty percent of all videotapes were coded by a second person for intercoder
reliability (r= .994, SD =. 005).
Results and Discussion
A repeated measures ANOVA with ground condition (within- and across-category) as the
between-subject variable and trials (salience and test) as the within-subject variable yielded
only a main effect of ground condition, F (1, 22) = 7.15, p = .014, η2= .25. No reliable
preference for either event emerged in the salience trial, t (23) = 1.74, p = .09, and no
condition by salience trial interaction was found, p >.21. Only infants in the across-category
condition preferred to look at the novel grounds at test, t (12) = 2.27, p = .04, and this
occurred at above chance level (see Figure 6). No significant difference was found for
different comparisons of grounds in each condition (within- and across-categories).
Another repeated-measures ANOVA with condition (within- vs. across-category) as the
between-subject variable and 4 familiarization trials as within-subject variables yielded only
a main effect of familiarization trial, F(3, 66)= 2.86, p= .04, ηp
2 = .12. No main effect of
condition or an interaction between familiarization trial and condition were found. Infants in
both conditions were equally attentive during the familiarization trials and their looking
times declined across trials.
Göksun et al. Page 13
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Thus, the results completely parallel those in Experiment 1b: when two grounds belonged to
different categories as defined in Japanese (such as the road vs. the grassy field comparison),
English-reared infants showed a preference for the novel ground on grayscale just as they
had in the color exposure.
We also compared the percentage of looking time to the novel ground between the infants
who watched the movies on the color screen in Experiment 1b and in grayscale. Results
showed that infants’ looking time patterns were very similar between the two studies with
no significant differences in looking time, p> .05 (see Figures 5 and 6).
The results of Experiment 1c argue against the possibility that infants were simply using the
color of the grounds as a feature for ground categorization. Although there might be other
perceptual confounds in the scene, the findings suggest that the geometry of the ground is a
strong cue for differentiating between two grounds on a color screen and in grayscale.
Hence, the nature of the ground, here defined abstractly as connecting specific starting and
ending points and extending in a line or a plane, is noticed by English-reared infants when
they process these dynamic events. Together with the findings from Experiment 1b, these
findings suggest that the perceptual differences among grounds fall into two distinct
categories, categories described by the Japanese verbs wataru and tooru.
Experiment 2: Discrimination of Figures and Grounds in Static Representations of Events
Verb learning demands perceiving the spatial-temporal interaction inherent in dynamic
events, because verbs capture a categorical moment in the unfolding event. For example,
consider a woman running on the street. The woman runs in space, on a ground (i.e., the
street) within a specific period of time. As she runs, both spatial and temporal dimensions
change. Motion verbs are defined in large part by the interaction of figures and grounds
across space and time. We thus wanted to explore whether infants perceive the semantic
components of events in static displays of the same dynamic events. These scenes preserve
the spatial dimension but not temporal aspect of those events. A picture of a dynamic event
takes “a slice in time” as if the temporal dimension is frozen while keeping a static spatial
configuration. Is the dynamic information important to the kinds of categories formed for
verb learning or would static slices in time equally preserve the dimensions reflected in verb
learning?
Past literature on containment, support, and degree of fit studies all utilized dynamic events
showing a hand that placed objects into particular spatial relationships (Casasola & Cohen,
2002; Choi, 2006; Hespos & Spelke, 2004). Yet, the crucial distinctions between these
events and ours are that in the past literature, discrimination of semantic elements could be
made based on the static endpoints of the events. For instance, “putting a toy tightly into a
box” would be presented as an actual hand moving a toy tightly into a box, resulting in a
static scene. In stark contrast, the ground-path verbs used in Japanese represent the
interaction of the path of the figure against a particular type of ground. The verb wataru, for
example, ‘flat barriers dividing two points’ from the dynamic ground discrimination study
implies a dynamic motion on a particular ground rather than just a representation of the
ground itself. It is possible that by teasing apart the spatial and temporal dimensions, static
events might no longer represent these scenes as different enough to be codified with
different verbs (Muehleisen & Imai, 1997).
No research (of which we are aware) has investigated infants’ ability to distinguish figures
and grounds in static representations of dynamic events. Only one published study examined
infants’ attention to the relation between figure and context by using pictures of animals and
vehicles on various grounds (Bornstein, Arterberry, & Mash, 2010). They found that 6month-old
infants categorized colored photographs of animals and vehicles across various
Göksun et al. Page 14
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
contexts in which the figures appeared (e.g., a tiger in the green field, a tiger on a beach, and
a tiger in a parking lot). The authors concluded that in category learning, infants initially
focus on objects and ignore the contextual information.
Experiment 2 explored whether eliminating the temporal, but preserving the spatial aspect of
an event enhanced or detracted from infants’ ability to distinguish different figures and
grounds. Instead of dynamic events, we used still shots of the people when they were in the
middle of the screen. A static display might thus reduce attention to the figure and give
priority to the ground, yielding an earlier discrimination of grounds. These studies allow us
to specify whether infants notice the same semantic components to the same degree in static
vs. dynamic events.
Eight- and 11-month-old infants’ processing of figures and grounds was tested in static
versions of the dynamic motion events used in Experiment 1a. These age groups were
chosen to parallel those in the dynamic figure and ground discrimination studies
(Experiment 1a). Also, given the findings of Bornstein et al. (2010), we expected that infants
at these ages might be able to detect changes in figures and ground in static pictures. We
expected that 1) there would be no a priori preference to static scenes before familiarization
to a specific figure or ground; 2) similar to the dynamic experiments, in general, infants
should look longer to the novel figure or the novel ground at test; 3) English-reared infants
might not be sensitive to different ground distinctions (i.e., within- vs. across-category
distinctions) in static events that no longer have a temporal dimension; and 4) infants might
detect grounds earlier than they did in dynamic events, due to the presence of a static figure
which might not attract as much attention as a dynamic figure.
Method
Participants
Thirty-five 8-month-old (M= 8.14, SD= .85, 18 males) and 32 11-month-old (M=11.12,
SD= .88, 13 males) English-reared infants participated. Infants were randomly assigned to
the figure discrimination study (17 8-month-olds and 16 11-month-olds) or the ground
discrimination study (18 8-month-olds and 16 11-month-olds). All infants were monolingual
and full-term at birth. Infants were predominantly White and from middle-class homes in
two Northeastern towns in the United States. An additional 22 infants across two age groups
and figure and ground studies were excluded from data analyses because they were bilingual
(n = 3), premature (n = 1), experimental error (n = 4), and having low attention to the video
clips (n = 12), or having a side bias (n =2).
Stimuli
This time, instead of dynamic events, we used screen shots of the figures (a woman, a man,
a six-year-old girl, and a six-year-old boy) when they were in the middle of the screen
(equally distant from both ends of the screen). All stimuli at all phases were static versions
of the dynamic events. No linguistic audio accompanied these scenes.
As in the dynamic figure discrimination study, there were three comparisons of figures:
adult-adult, adult-child, and child-child. The figures were also presented on different
grounds within in each condition. For grounds, we again preserved two conditions based on
how Japanese would encode these ground verbs in dynamic contexts. One condition
involved the comparison of two grounds from the category of wataru (e.g., a road vs. a
railroad track) and the other condition included cross-category comparisons with on ground
from the “wataru” category and another from the “tooru” category (e.g., a road vs. a tennis
court).
Göksun et al. Page 15
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Procedure and Coding
The procedure and coding for the static figure and ground discrimination studies were
exactly the same as the dynamic figure and ground studies. Static stimuli were presented for
the same amount of time as the dynamic stimuli. A second person coded 20% of all
videotapes for inter-coder reliability (r= .994, SD= .003).
Results
We calculated infants’ percentage of looking time towards each scene in the split-screen for
salience and test trials. We report results separately for the figure and ground studies. No
significant gender differences emerged for each study. Thus, gender was not considered as a
separate factor for further analyses.
Figure discrimination
A repeated-measures ANOVA with age (8- and 11-month-olds) as the between-subject
variable and trials (salience and test) as the within-subject variable yielded main effects of
age and trial, F (1, 31) = 8.91, p= .01, η2= .22 and F (1, 31) = 8.12, p= .01, η2= .20, but no
significant interaction between them. Infants’ looking times to the left and right side of the
screen did not significantly differ in the salience trials at either age: 8-month-olds: t (16) = .
99, p = .34, 11-month-olds: t (15) = 1.06, p = .31. As shown in Figure 7 (top panel), only 11month-old
infants looked significantly longer to the novel figure compared to the same
figure at test, t (15) = 3.03, p = .01. Infants in both age groups were equally attentive to the
whole movie (63% and 66%, respectively for 8- and 11-month-olds). Looking times during
the familiarization phase yielded no main effects of familiarization trial, age, or an age by
familiarization trial interaction (Fs < 1.98, ps > .08).
Ground discrimination
A repeated-measures ANOVA with age (8- and 11-month-olds) as the between-subject
variable and trials (salience and test) as the within-subject variable yielded a main effect of
trial, F (1, 32) = 8.38, p= .01, η2= .21 and a trial by age interaction, F (1, 32) = 6.03, p= .02,
η2= .16. That is, infants at both age groups had no significant preference for either event at
any age in the salience trial: 8-month-olds: t (17) = .59, p = .56, 11-month-olds: t (15) = .93,
p = .37. Only 11-month-old infants looked significantly longer to the novel ground
compared to the familiar ground at test, t (15) = 5.45, p = .01 (see Figure 7, bottom panel).
Only the older age group (11-month-olds) appeared to notice the change in the ground in the
test trial after familiarization, though the age groups were similar in their total attention
during the whole movie (66% and 64%, respectively for 8- and 11-month-olds).
Interestingly, and as predicted, infants detected the grounds in static scenes earlier than they
noted ground changes in dynamic displays (Experiment 1b). Looking times during the
familiarization phase yielded no main effects of familiarization trial, age, or an age by
familiarization trial interaction (Fs < .84, ps > .44).
Next, we analyzed whether infants were sensitive to the categorical ground distinctions
coded by Japanese in static representations that lacked a temporal component. In stark
contrast to what we found in Experiments 1b and 1c, infants started noticing the changes of
grounds in static representations at 11 months of age and looked similarly to novel grounds
regardless of the categorical distinction in Japanese. Also, no significant difference was
found for different comparisons of grounds in each condition, p> .05 (within- and across-
categories).
Finally, we tested infants’ looking times during familiarization in the figure and ground
studies for the dynamic versus the static displays (the comparison of Experiments 1a and 2).
Göksun et al. Page 16
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Results indicated that infants looked significantly longer to the dynamic displays than to the
static displays for both figures and grounds, F(1, 76) = 9.37, p = .003, η2 = .11, and F(1, 78)
= 13.96, p = .001, η2 = .15. No significant interaction between age and study (dynamic vs.
static) was found, suggesting that dynamic displays were more interesting to infants
compared to static scenes.
Discussion
Experiment 2 demonstrated that English-reared infants differentiated static figures by 11
months of age, which was the same time they detected moving figures in dynamic motion
events. In contrast, ground discrimination was detected earlier when the stimuli were static.
In particular, infants noticed the change in grounds in static events 3 months before they did
so for dynamic events. On the other hand, in the static versions of the dynamic events,
infants did not distinguish between the types of grounds coded in Japanese. They treated
within- and across-category comparisons of grounds similarly.
This experiment suggests that infants do not process static and dynamic events in the same
way (Cutting & Profitt, 1981). A critical finding, for example, was the disappearance of the
categorical Japanese ground distinctions in the static versions of the same dynamic events.
Without the dynamic motion, infants differentiated grounds both within- and acrosscategories,
indicating that the perceptual distinctions between grounds in across-category
comparisons were not more salient than within-category comparisons. But static scenes fail
to capture the interaction between the figure and ground evident in dynamic scenes and
therefore do not support the categorization of grounds. Arguably, the movement of the figure
in dynamic events clarifies the starting and ending points of the grounds and whether the
ground extends in a line or in a plane. The results also suggest that understanding verbs
demands actual interaction between figures and grounds in actual events.
The removal of the dynamic aspects of the event also had the effect of enhancing infants’
ability to distinguish between different grounds. Presumably, the person’s movement in
Experiment 1 oriented infants’ attention to figures rather than grounds. Movement is a
strong predictor of infants’ attention to figures (e.g., Kellman & Spelke, 1983; Otsuka &
Yamaguchi, 2003). Thus, younger infants appear to experience change-blindness as to
“where” the action occurs when there are moving figures. Interestingly, during
familiarization, infants had similar looking times in the figure discrimination and ground
discrimination conditions for the static scenes, which were significantly lower than the
looking times in the dynamic displays (which were also similar for the figure and ground
condition).
Finally, this result raises questions about why infants successfully distinguish animals and
vehicles in various environments at 6 months of age in an earlier study (Bornstein et al.,
2010), but detect static and dynamic human figures later. One reason that processing might
be more difficult in this study in relation to Bornstein et al. is that we presented the humans
using a lateral view of the figures on the screen, which made it more difficult to notice faces
which might otherwise be more inherently interesting. Another explanation might be the size
and location of the objects: the people were in the middle of the screen and the grounds
occupied more space in comparison to the size of the figures in the static events.
Experiment 3: Cross-Language Comparisons on the Detection of Grounds
Our findings on infants’ processing of grounds in dynamic events demonstrate that Englishreared
infants notice non-native ground distinctions. However, the evidence up to this point
is inconclusive about the role language plays in infants’ processing of these categorical
ground distinctions. Would Japanese-reared infants process grounds in dynamic events in
Göksun et al. Page 17
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
the same way as that evidenced by their English-reared peers? If so, it would offer support
for the concept-to-language hypothesis. Testing Japanese infants also enables us to ask
whether the early categorical distinctions infants notice in events change as they are exposed
to their ambient language. Japanese children, immersed in a language that codes for these
ground distinctions, should continue to notice different classes of grounds (e.g., wataru and
tooru) as they become speakers of their language. English-reared infants, who do not hear
these distinctions marked in their ambient language, should eventually dampen their ability
to detect these event components.
This work parallels previous research demonstrating that children’s sensitivity to non-native
categorical distinctions decreases over time (e.g., Choi & Bowerman, 1991; Choi, 2006;
Gentner & Bowerman, 2009). For example, after the second year of life, English- and
Korean-speaking children responded differently to Korean tight- vs. loose-fit distinctions,
suggesting that language-specific aspects of these spatial categories influence children’s
nonlinguistic sensitivity. Thus, learning language dampens the detection of non-linguistic
categorical differences that are not encoded in one’s native tongue. In the final set of
experiments we asked whether Japanese-reared infants differentiate between grounds in a
way similar to English-reared infants and whether the early categorical distinctions infants
notice in events change with exposure to their ambient language.
Experiment 3a: Can 14-Month-Old Japanese-reared Infants Discriminate Between
Grounds?
Here we examined how Japanese-reared infants processed grounds in dynamic events and
whether they were receptive to the categorical ground distinctions encoded in Japanese.
Fourteen-month-old Japanese-reared infants were tested because infants from Englishspeaking
environments differentiate between grounds at this age. First, it was hypothesized
that there would be no preference for either event during the split-screen salience trials.
Second, if infants discriminate between grounds in the test trial, they will look longer to the
side where there is a novel ground. Last, similar to English-reared infants’ responses,
Japanese-reared infants would also differentiate grounds better when the comparison was
between two categories in Japanese (wataru vs. tooru) as opposed to within the same
Japanese verb category (wataru).
Method
Participants
The final sample was comprised of 26 14-month-old Japanese-reared infants (M= 14.07,
SD= .74, 15 males). Data were collected in the Greater-Tokyo Metropolitan area Japan by a
trained Japanese-English bilingual experimenter. All infants were full-term and came from
monolingual Japanese households. An additional 5 infants were excluded from data analyses
because they fussed-out during the experiment (n = 4) or had low attention to the video clips
(n = 1).
Stimuli, Procedure, and Coding
All were the same as those used in the dynamic ground discrimination study of Experiment
1. A second coding of 12% of subjects yielded high reliability (r= .992, SD = .003).
Results and Discussion
A repeated measures ANOVA with ground condition (within- and across-category) as the
between-subject variable and trials (salience and test) as the within- subject variable yielded
only main effects of ground condition and trial, F (1, 24) = 16.32, p = .001, η2= .41 and F
Göksun et al. Page 18
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
(1, 24) = 7.37, p = .012, η2= .24, respectively. The interaction between trial and ground
condition was marginally significant, F (1, 24) = 3.83, p = .06, η2= .14. No significant
preference for either event emerged in the salience trial in either condition, ts < 1.83, ps > .
09. There was also no significant condition by salience trial interaction, p > .14. Only infants
in the across-category condition preferred to look at the novel grounds at test, t (10) = 7.75,
p = .001, and this result was above chance level (see Figure 8, bottom panel). Another
repeated-measures ANOVA with condition (within- vs. across-category) as the betweensubject
variable and 4 familiarization trials as within-subject variables yielded only a main
effect of familiarization trial, F(3,72)= 4.44, p= .006, ηp
2 = .16. Neither a main effect of
condition (within- vs. across-category) nor an interaction between trial and condition were
found. Infants in both conditions were equally attentive during the familiarization trials and
looking times declined across trials.
Finally, we compared English-reared infants from Experiment 1b and Japanese-reared
infants in this experiment. Results indicated no main effect of language group or any
interactions with language group, Fs < 1.09, ps > .30 (see Figure 8). Both age groups
significantly differentiated between grounds only in the across-category conditions.
The findings of this study suggest that 14-month-old Japanese infants and English-reared
children of the same age demonstrated very similar sensitivity to distinctions between
grounds in dynamic events. Just as English- and Korean-reared infants both detect the
degree-of-fit relations in containment and support events even though this information was
encoded only in Korean (Choi, 2006), English- and Japanese-reared 14-month-olds both
attended to ground distinctions in nonlinguistic events in ways that are consistent with the
category cuts used in Japanese.
These results support the concept-to-language hypothesis as well as Gentner and
Bowerman’s (2009) conjecture that some event categories might be acquired before
language learning. An accurate assessment of these theories cannot be complete without
testing the role of language learning on these event constructs. In the final experiment, we
examine both English- and Japanese-reared older children’s attention to the ground
categorical distinctions in dynamic, nonlinguistic events.
Experiment 3b: Can 19-Month-Old English- and Japanese-reared discriminate between
grounds similarly?
Here we explore the link between learning a native language (i.e., English or Japanese) and
the processing of grounds in nonlinguistic events. If infants’ attention to events alters as they
learn their ambient language, we might expect differences between English- and Japanesereared
toddlers. Nineteen-month-old English- and Japanese-reared toddlers who generally
pass the 50-word mark were recruited as participants and compared to those who were at the
cusp of language learning in our prior experiments. At around 18 months of age, children
seem to undergo a notable increase in the amount of vocabulary produced (e.g., L. Bloom,
1973; Dromi, 1987; Gopnik & Meltzoff, 1987; but see P. Bloom, 2000). Therefore, this is a
good age at which to test the effect of language environment on processing nonlinguistic
event categories.
We predict that Japanese children, immersed in a language that uses ground distinctions
(wataru ‘flat barriers dividing two points’ vs. tooru ‘continuous plane’), should maintain the
distinction between these classes of grounds. As before, if toddlers discriminate between
grounds in the test trial, they will look longer to the side where there is a novel ground.
Göksun et al. Page 19
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Method
Participants
Twenty-four 19-month-old (M=19.15, SD= 1.02, 13 males) English-reared and 24 Japanesereared
toddlers (M=19.28, SD= 1.10, 11 males) from monolingual homes were tested. An
additional 10 infants (17%) across both groups were excluded because they were bilingual
(n = 1), fussed-out (n = 7), had low attention (n = 1) or a side bias (n =1).
Stimuli, Procedure, and Coding
All were exactly the same as the dynamic ground discrimination study of Experiments 1 and
3a. Inter-coder reliability on 12% of subjects was high (r = .995, SD = .004).
Results and Discussion
Results from the familiarization phase showed that both language groups and conditions
(within- vs. across-category) were attentive. Looking times declined during familiarization
across the trials, F(3, 132)= 8.48, p= .01, ηp
2 = .16.
A repeated measures ANOVA using language group and condition as the between-subject
variables and two trials (salience and test) as the within-subject variables was calculated.
There were main effects of language group and condition only on test trials, F(1, 44)= 4.34,
p= .04, ηp
2 = .09 and F(1, 44)= 8.42, p= .01, ηp
2 = .16 (see Figure 8) and no significant
language group by condition interaction emerged. Neither English- reared nor Japanesereared
toddlers preferred either event in the salience trials, ts < .84, ps > .41.
Planned pair-wise comparisons suggest that only Japanese-reared toddlers who were in the
across-category condition looked significantly longer to the novel ground in the change
trials, t(11)= 5.19, p = .01 (see Figure 8, bottom panel). Japanese-reared children in the
within-category condition and English-reared children in both conditions looked equally to
both sides at test, ts < 1.72, ps > .11.
Here, we predicted that only Japanese children, immersed in a language that supports ground
distinctions, would maintain those distinctions whereas English-reared toddlers might
dampen attention to contrasts not encoded in their native language. The results confirmed
these predictions. Even though 14-month-olds from both English and Japanese language
environments were equally sensitive to distinctions that the Japanese language makes among
grounds in dynamic events (Experiments 1b and 3a), this effect was dampened for 19month-old
English-reared toddlers. These findings were not the consequence of general
attentional differences between the language groups. One alternative possibility is that
English 19-month-olds’ lack of preference between grounds is a function of attentional
biases toward the focal object/agent that have been documented in English-speaking cultures
in adults (e.g., Masuda & Nisbett, 2001, 2005). In other words, infants would not attend to
ground features sufficiently to make the distinctions between ground categories. Future
studies using eye-tracking methods will address this question. The current study offers
preliminary evidence that language plays a role in the way in which we process events; if
ground distinctions are not made linguistically (as in English) children’s ability to detect
these distinctions falters.
General Discussion
Figure and ground are foundational elements in events and are codified across languages.
Whether you speak English or Turkish or Japanese, languages comment on events by
frequently expressing the figure in the event and sometimes expressing the ground. Like
Göksun et al. Page 20
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
other event components, figure and ground are perceptually accessible, universally
recognized, and codified differently across languages (Golinkoff & Hirsh-Pasek, 2008). The
current research expands the literature by studying the roots of these semantic distinctions in
infant perception. In particular, we asked whether and how very young children perceive
figures and grounds in events and how this perception might be modified when children start
learning their native language.
Our findings indicated that, first, English-reared infants noticed changes in figures and
grounds in dynamic events (Experiments 1a and 1b). Notably, infants were receptive to the
non-native categorical ground distinctions for crossing action only in dynamic events. That
is, the Japanese distinction proved to be a “default” for English-reared infants (Experiment
1b). Second, the geometry of the ground appeared to be a strong cue to distinguish between
two grounds on both a color screen and in grayscale (Experiments 1b and 1c). Third, even
though English-reared infants detected figures and grounds in static representations of the
dynamic crossing events, the Japanese categorical ground differentiation no longer emerged
in static displays, indicating that in these displays within- and across-category comparisons
were treated similarly (Experiment 2). Finally, the early sensitivity to categorical ground
distinctions by English-and Japanese-reared infants diverged as children began to process
language patterns in their native languages. Japanese, but not English-reared toddlers
preserved these distinctions, suggesting that the process of learning language appears to shift
early-formed categorical boundaries (Experiment 3). Together, the current experiments
present evidence that infants parse nonlinguistic dynamic events (and their static
representations) into components to detect “who is doing the action where.”
Our findings on figure and ground detection in dynamic events add to the growing literature
on infants’ perception of the foundational components of events (e.g., Casasola & Cohen,
2002; Lakusta et al., 2007; Pulverman et al., 2008), suggesting that by the latter half of the
first year and the beginning of the second year, infants attend to nonlinguistic event
components that are represented across languages. This is an essential first step in learning
relational terms, particularly for motion verbs (Gentner & Bowerman, 2009; Göksun et al.,
2010; Golinkoff & Hirsh-Pasek, 2008). Yet, some of these concepts seem to surface later in
development than others (e.g., noticing figure changes before ground changes in dynamic
events).
Once these conceptual foundations are present, they can be combined to create the semantic
bases for word-to-world relations. For example, the early detection of the figures in dynamic
events (e.g., Golinkoff & Kerr, 1978; Oakes, 1994; Robertson & Suci, 1980) allows children
to name them early on (Clark, 1979; Grace & Suci, 1981; Tomasello & Merriman, 1995).
One question is what truly constitutes a “figure” from an infant’s point of view. The
presence of people in a scene might enhance infants’ attention, and thus their ability to
notice changes. But will children have similar reactions to different figures such as cars,
animals, and balls? Because the verb ‘crossing’ is permitted for all types of figures (e.g., the
ball crossed, the person crossed, the dog crossed), children should respond similarly to
different figure types in nonlinguistic events and extend their verbs to various figures.
Future studies should tease apart the role of animacy in noticing the figures in different
scenes.
Another finding was that changes in figures and grounds are differentially recognizable in
static and dynamic displays. When the dynamic aspects of the events are removed, in
general, infants’ ability to differentiate between grounds is improved. Nevertheless, the verb
wataru, for example, ‘flat barriers dividing two points’ implies a dynamic motion on a
particular ground rather than just a representation of the ground itself. Dynamic information
with the attendant spatial and temporal interaction appears to be necessary to reflect
Göksun et al. Page 21
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
“categories” for verb learning. A slice in time without the temporal component of an event is
not enough to maintain categorical ground-path distinctions in events. In addition,
preliminary results from a further study in which the path of the motion was modified from
crossing to walking alongside/near the grounds, indicate that 14-month-old infants do not
differentiate between grounds based on categorical distinctions when figures were walking
near and not across grounds. This confirms the findings from the static and dynamic ground
studies, suggesting that categorical ground distinctions occur only with the relevant
interactions by a figure with ground and path (Göksun, 2010).
Even adults have difficulty detecting obvious changes in a picture of a scene when two
pictures are presented sequentially (Simons & Levine, 1997). Nevertheless, cross-cultural
differences appear in the detection of changes in scenes (Ji, Peng, & Nisbett, 2000, 2004;
Masuda & Nisbett, 2001, 2005). Nisbett and colleagues investigated East Asians’ (e.g.,
Japanese, Chinese, and Taiwanese) and North Americans’ attention to the context and
figures in a scene. East Asians were more attentive to relationships between objects and their
environment than North Americans (Ji, Peng, & Nisbett, 2000; Ji, Zhang, & Nisbett, 2004).
Masuda and Nisbett (2001) also showed that compared to Americans, Japanese-speaking
adults described the background of an underwater scene and expressed more relationships
between the focal figure (e.g., a fish) and the background. In another study, using the change
blindness paradigm (i.e., failure to detect the changes in a scene after exposure to it),
Masuda and Nisbett (2005) displayed two animated scenes (e.g., a farm) that differed in
small details. American adults again detected changes in the focal objects, but Japanese
adults noticed the changes in the context and the relationships between the objects. Although
these findings were mainly discussed in relation to cultural variations, the evidence
corroborates our results on the difference between Japanese and American toddlers’
differentiation of grounds in dynamic events. Thus, these findings on change blindness
could also be interpreted as a consequence of language rather than only cultural differences.
Future studies are necessary to disambiguate these two interpretations.
Conceptual Distinctions Before and After Language
Despite differences in the ways in which languages encode foundational event components,
infants seem to process these event constructs similarly before language has a chance to
influence their perception (Hespos & Spelke, 2004; 2007; Göksun et al., 2010; McDonough,
Choi, & Mandler, 2003; Mandler, 2004). The ground discrimination results with dynamic
events support the claim that early event perception might be universal. English-reared
infants are sensitive to the nature of the ground in a way that is more specific than encoded
by the English verb ‘cross.’ This is consistent with findings from Hespos and Spelke (2004)
that English-reared infants’ discriminate degree-of-fit relationships between two objects that
are not encoded in English. In the current study, a conceptual distinction for grounds in
dynamic events was revealed, demonstrating that infants noticed differences when a ground
extends in a line (e.g., road) or extends in a plane (e.g., grassy field). That is, irrespective of
the language environment in which infants are raised, they detect non-linguistic components
of events, and infants attend to fine-grained distinctions in events even when these are not
codified in their native language. The conceptual underpinnings for the learning of a
language’s relational terms are in place as proposed by concept-to-language hypothesis
(Gentner & Bowerman, 2009). As children note how these event components are lexicalized
in their native tongue, they appear to tune into certain semantic distinctions over others,
influenced by the ambient language. Future studies with English-speaking adults are planned
to address the question of how readily these ground distinctions can be resurrected.
These findings are reminiscent of, though not analogous to, the universal phonological
categories prelinguistic infants possess (e.g., Eimas, Miller, & Jusczyk, 1987; Kuhl et al.,
1997; Werker & Tees, 1984). There might be a broad set of foundational components in
Göksun et al. Page 22
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
events that will later be dampened by attending to only the subset that are coded in one’s
native language. While learning language, children might semantically reorganize their
prelinguistic constructs by either subtly dividing their spatial and temporal world further or
by creating a broader category (for details of this argument see Göksun et al., 2010; Hespos
& Spelke, 2007).
A possibility that emerges from these findings carries implications for the learning of
relational terms like verbs and prepositions in special populations. Perhaps problems in the
learning of relational terms result partially from the inability to either perceive event
components or reorganize event components with native language exposure. Children with
autism, for example, might have particular problems with the processing of figures, because
they show less attention to people in their environment. In contrast, children with autism
might find some figures such as cars to be intrinsically more interesting than people (Sasson,
Turner-Brown, Holtzclaw, Lam, & Bodfish, 2008). Thus, children with autism might
respond differentially to the same event with different types of figures.
There is still much to explore on infants’ ability to notice and abstract spatial and event
components that are fundamental for learning verbs and prepositions. By bridging linguistics
and event perception, we explored how infants process two components of events, figure
and ground. We also selected perceptually salient ground-path categories. To generalize our
conclusions, other grounds that are incorporated in verbs should be examined. More crosslinguistic
studies - as well as studies with bilingual children - are required to confirm our
assertions about the universal to language-sensitive shift. Studies across typologically varied
languages about how children acquire the biases of their native language will illuminate the
developmental links between language and thought.
Conclusions
Our findings begin to reconcile the long-standing debate on the role of language in shaping
cognition, and provide support for concept-to-language hypothesis at least for very central
and highly perceptual events constructs. That is, children’s initial perception of events is not
a ‘kaleidoscopic flux of impressions” as suggested by Whorf (1956) that awaits language for
its organization. Rather, infants by 14 months are forming categories of the components of
nonlinguistic events that appear to be the same regardless of the ambient language.
Language appears to play a role in event perception once children notice how language
‘underlines’ different aspects of events.
Acknowledgments
This work was supported by NICHD grant 5R01HD050199 and by NSF grants BCS-0642529 to the second and
third authors, Japan Ministry of Education KAKENHI grant 18300089 and Tamagawa University Global Center of
Excellence (GCOE) program to the fourth and last author, respectively. Portions of this research have been
presented at the 16th International Conference on Infant Studies, at the 11th Meeting of International Association
for the Study of Child Language, and at the 33rd and 34th Meetings of Boston University Conference on Language
Development. We would like to thank everyone at the Temple University Infant Lab, the University of Delaware
Infant Language Project, Keio University Imai Lab, and Tamagawa University Baby Lab for their invaluable
contributions to this project. Special thanks to Nora Newcombe for insightful comments on each study, Sarah
Roseberry for discussions on many issues about the studies, Wendy Shallcross, Yannos Misitzis, Katrina Ferrara,
Russell Richie, and Aimee Stahl for their help in data collection. We would also like to express our deepest
appreciation to all of the parents and infants who participated in the study.
References
Aguiar A, Baillargeon R. 2.5-month-old infants’ reasoning about when objects should and should not
be occluded. Cognitive Psychology. 1999; 39:116–157. [PubMed: 10462457]
Göksun et al. Page 23
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Bahrick LE, Newell LC. Infant discrimination of faces in naturalistic events: Actions are more salient
than faces. Developmental Psychology. 2008; 44:983–996. [PubMed: 18605829]
Baillargeon R, Needham A, DeVos J. The development of young infants’ intuitions about support.
Early Development and Parenting. 1992; 1:69–78.
Baldwin DA, Baird JA, Saylor MM, Clark AM. Infants parse dynamic action. Child Development.
2001; 72:708–717. [PubMed: 11405577]
Beavers J. On the Nature of Goal Marking and Event Delimitation: Evidence from Japanese. Journal
of Linguistics. 2008; 44:283–316.
Bertenthal, BI. Infants’ perception of biomechanical motions: Intrinsic image and knowledge-based
constraints. In: Granrud, C., editor. Visual perception and cognition in infancy. Erlbaum; Hillsdale,
NJ: 1993. p. 175-214.
Bloom, L. One word at a time: The use of single-word utterances before syntax. The Hague; Mouton:
1973.
Bloom, P. How children learn the meanings of words. MIT Press; Cambridge, MA: 2000.
Bock KJ, Warren RK. Conceptual accessibility and syntactic structure in sentence formulation.
Cognition. 1985; 21:47–67. [PubMed: 4075761]
Bogartz RS, Shinskey JL, Schilling TH. Object permanence in five-and- a-half-month-old infants.
Infancy. 2000; 1:403–428.
Bornstein MH, Arterberry ME, Mash C. Infant object categorization transcends object-context
relations. Infant Behavior and Development. 2010; 33:7–15. [PubMed: 20031232]
Bowerman, M. Containment, support, and beyond: Constructing topological spatial categories in first
language acquisition. In: Aurnague, M.; Hickmann, M.; Vieu, L., editors. The categorization of
spatial entities in language and cognition. John Benjamins; Amsterdam: 2007. p. 177-203.
Bowerman, M.; Choi, S. Shaping meanings for language: Universal and language specific in the
acquisition of spatial semantic categories. In: Bowerman, M.; Levinson, SC., editors. Language
acquisition and conceptual development. Cambridge University Press; New York, NY: 2001. p.
475-512.
Bowerman, M.; Levinson, SC., editors. Language acquisition and conceptual development. Cambridge
University Press; Cambridge: 2001.
Casasola M, Cohen LB. Infant categorization of containment, support, and tight-fit spatial
relationships. Developmental Science. 2002; 5:247–264.
Choi S. Influence of language-specific input on spatial cognition: Categories of containment. First
Language. 2006; 26:207–232.
Choi S, Bowerman M. Learning to express motion events in English and Korean: The influence of
language-specific lexicalization patterns. Cognition. 1991; 41:83–121. [PubMed: 1790656]
Choi-Jonin, I.; Sarda, L. The expression of semantic components and the nature of ground entity in
orientation motion verbs: A cross-linguistic account based on French and Korean. In: Aurnague,
M.; Hickman, M.; Vieu, L., editors. The categorization of spatial entities in language and
cognition. Benjamin Publishers; Amsterdam, Netherlands: 2007. p. 123-149.
Clark, EV. Building a vocabulary: Words for objects, actions and relations. In: Fletcher, P.; Garman,
M., editors. Language acquisition. Cambridge University Press; Cambridge: 1979. p. 149-160.
Clark, EV. First language acquisition. Cambridge University Press; New York: 2003.
Clark, EV. First language acquisition. 2nd edition. Cambridge University Press; New York: 2009.
Cutting, JE.; Profitt, DR. Gait perception as an example of how we may perceive events. In: Walk, R.;
Pick, HL., editors. Intersensory perception and sensory integration. Plenum; New York: 1981. p.
249-273.
Dromi, E. Early lexical development. Cambridge University Press; Cambridge, England: 1987.
Eimas PD. Categorization in early infancy and the continuity of development. Cognition. 1994; 50:83–
93. [PubMed: 8039377]
Eimas, PD.; Miller, JL.; Jusczyk, PW. On infant speech perception and the acquisition of language. In:
Harnard, S., editor. Categorical perception: The groundwork of cognition. Cambridge University
Press; 1987. p. 161-198.
Göksun et al. Page 24
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Fenson L, Dale PS, Reznick JS, Bates E. Variability in early communicative development.
Monographs of the Society for Research in Child Development. 1994; 59
Fenson L, Pethick S, Renda C, Cox JL, Dale PS, Reznick JS. Short-form versions of the MacArthur
Communicative Development Inventories. Applied Psycholinguistics. 2000; 21:95–116.
Gentner, D. Why nouns are learned before verbs: Linguistic relativity versus natural partitioning. In:
Kuczaj, S., editor. Language development: Language, thought, and culture. Vol. 2. Lawrence
Erlbaum Associates; Hillsdale, NJ: 1982. p. 301-334.
Gentner, D.; Bowerman, M. Why some spatial semantic categories are harder to learn than others: The
typological prevalence hypothesis. In: Guo, J.; Lieven, E.; Ervin-Tripp, S.; Budwig, N.;
Özçaliskan, S.; Nakamura, K., editors. Crosslinguistic approaches to the psychology of language:
Research in the tradition of Dan Isaac Slobin. Lawrence Erlbaum Associates; NJ, New York:
2009. p. 465-480.
Göksun, T. Unpublished dissertation. Temple University; Philadelphia, PA: 2010. The ‘who’ and
‘where’ of events: Infants’ processing of figures and grounds in nonlinguistic events.
Göksun T, Hirsh-Pasek K, Golinkoff RM. Trading Spaces: Carving up events for learning language.
Perspectives on Psychological Science. 2010; 5:33–42.
Golinkoff RM. Semantic development in infants: The concepts of agent and recipient. Merrill-Palmer
Quarterly. 1975; 21:181–193.
Golinkoff RM. The case for semantic relations: evidence from the verbal and nonverbal domains.
Journal of Child Language. 1981; 8:413–437. [PubMed: 7251715]
Golinkoff RM, Hirsh-Pasek K. How toddlers begin to learn verbs. Trends in Cognitive Sciences. 2008;
12:397–403. [PubMed: 18760656]
Golinkoff RM, Kerr JL. Infant’s perception of semantically defined action role changes in filmed
events. Merrill-Palmer Quarterly. 1978; 24:53–61.
Golinkoff RM, Hirsh-Pasek K, Cauley KM, Gordon L. The eyes have it: Lexical and syntactic
comprehension in a new paradigm. Journal of Child Language. 1987; 14:23–45. [PubMed:
3558524]
Gopnik A, Meltzoff AN. The development of categorization in the second year and its relation to other
cognitive and linguistic developments. Child Development. 1987; 58:1523–1531.
Grace, J.; Suci, GJ. The role of attentional priority of the agent in the acquisition of word reference;
Paper presented at the Society for Research on Child Development; Boston, MA. 1981;
Haith, MM. Rules that babies look by: The organization of newborn visual activity. Lawrence Erlbaum
Associates; Potomac, Maryland: 1980.
Hespos SJ, Baillargeon R. Reasoning about containment events in very young infants. Cognition.
2001a; 78:207–245. [PubMed: 11124350]
Hespos SJ, Baillargeon R. Infants’ knowledge about occlusion and containment events: A surprising
discrepancy. Psychological Science. 2001b; 12:141–147. [PubMed: 11340923]
Hespos SJ, Baillargeon R. Decalage in infants’ reasoning about occlusion and containment events:
Converging evidence from action tasks. Cognition. 2006; 99:B31–B41. [PubMed: 15939414]
Hespos SJ, Baillargeon R. Young infants’ actions reveal their developing knowledge of support
variables: Converging evidence for violation-of-expectation findings. Cognition. 2008; 107:304–
316. [PubMed: 17825814]
Hespos SJ, Saylor M, Grossman S. Infants’ ability to parse continuous actions series. Developmental
Psychology. 2009; 45:575–585. [PubMed: 19271840]
Hespos SJ, Spelke ES. Conceptual precursors to language. Nature. 2004; 430:453–456. [PubMed:
15269769]
Hespos, SJ.; Spelke, ES. Precursors to spatial language: The case of containment. In: Aurnague, M.;
Hickman, M.; Vieu, L., editors. The categorization of spatial entities in language and cognition.
Benjamin Publishers; Amsterdam, Netherlands: 2007. p. 233-245.
Hirsh-Pasek, K.; Golinkoff, RM. The origins of grammar: Evidence from early language
comprehension. MIT Press; Cambridge, MA: 1996.
Jackendoff, R. Semantics and cognition: Current studies in linguistics series, No. 8. The MIT Press;
Cambridge, MA: 1983.
Göksun et al. Page 25
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Jackendoff, R. Languages of the mind. MIT Press; Bradford: 1992.
Ji L, Peng K, Nisbett RE. Culture, control, and the relationships in the environment. Journal of
Personality and Social Psychology. 2000; 78:943–955. [PubMed: 10821200]
Ji L, Zhang Z, Nisbett RE. Is it culture or is it language? Examination of language effects in crosscultural
research on categorization. Journal of Personality and Social Psychology. 2004; 87:57–65.
[PubMed: 15250792]
Johnson SC. The recognition of mentalistic agents in infancy. Trends in Cognitive Sciences. 2000;
4:22–28. [PubMed: 10637619]
Johnson SP, Aslin RN. Perception of object unity in 2-month-old infants. Developmental Psychology.
1995; 31:739–745.
Johnson SP, Aslin RN. Perception of object unity in young infants: The roles of motion, depth, and
orientation. Cognitive Development. 1996; 11:161–180.
Johnson SP, Aslin RN. Young infants’ perception of illusory contours in dynamic displays. Perception.
1998; 27:341–353. [PubMed: 9775316]
Johnson SP, Mason U. Perception of illusory contours by 2-month-old infants. Child Development.
2002; 73:22–34. [PubMed: 14717241]
Kaufman-Hayoz R, Kaufman F, Stucki M. Kinetic contours in infants’ visual perception. Child
Development. 1986; 57:353–358.
Kellman PJ, Spelke ES. Perception of partly occluded objects in infancy. Cognitive Psychology. 1983;
15:483–524. [PubMed: 6641127]
Kimchi R, Peterson MA. Figure-ground Segmentation can occur without attention. Psychological
Science. 2008; 19:660–668. [PubMed: 18727781]
Koffka, K. Principles of Gestalt psychology. Harcourt Brace Jovanovich; New York: 1935.
Kuhl PK. Early language acquisition: Cracking the speech code. Nature Reviews Neuroscience. 2004;
5:831–843.
Kuhl PK, Andruski JE, Chistovich IA, Chistovich LA, Kozhevnikova EV, Ryskina VL. Crosslanguage
analysis of phonetic units in language addressed to infants. Science. 1997; 277:684–686.
[PubMed: 9235890]
Kuhl PK, Conboy BT, Padden D, Nelson T, Pruitt J. Early speech perception and later language
development: Implications for the “critical period.”. Language Learning and Development. 2005;
1:237–264.
Lakoff, G. Women, fire, and dangerous things: What categories reveal about the mind. University of
Chicago Press; Chicago: 1987.
Lakusta L, Wagner L, O’Hearn K, Landau B. Conceptual foundations of spatial language: evidence for
a goal bias in infants. Language Learning and Development. 2007; 3:179–197.
Landau B, Jackendoff R. “What” and “where” in spatial language and spatial cognition. Behavioral
and Brain Sciences. 1993; 16:217–238.
Langacker, RW. Foundations of cognitive grammar. Stanford University Press; Stanford, CA: 1987.
Maguire MJ, Hirsh-Pasek K, Golinkoff RM, Haryu E, Imai M, Vengas S, Okada H, Pulverman R,
Sanchez-Davis B. A developmental shift from similar to language-specific strategies in verb
acquisition: A comparison of English, Spanish and Japanese. Cognition. 2010; 114:299–319.
[PubMed: 19897183]
Malt, BC.; Wolff, P. Words and the mind: How words encode human experience. Oxford University
Press; New York: 2010.
Mandler JM. How to build a baby II: Conceptual primitives. Psychological Review. 1992; 99:587–
604. [PubMed: 1454900]
Mandler, JM. The foundations of mind: Origins of conceptual thought. Oxford University Press; New
York: 2004.
Mandler JM. The spatial foundations of the conceptual system. Language and Cognition. 2010; 2:21–
44.
Masuda T, Nisbett RE. Attending holistically versus analytically: Comparing the context sensitivity of
Japanese and Americans. Journal of Personality and Social Psychology. 2001; 81:922–934.
[PubMed: 11708567]
Göksun et al. Page 26
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Masuda T, Nisbett RE. Culture and change blindness. Cognitive Science. 2005; 30:381–399.
[PubMed: 21702819]
McDonough L, Choi S, Mandler JM. Understanding spatial relations: Flexible infants, lexical adults.
Cognitive Psychology. 2003; 46:229–259. [PubMed: 12694694]
Muehleisen, V.; Imai, M. Transitivity and the incorporation of ground information in Japanese path
verbs. In: Lee, K.; Sweetwer, E.; Verspoor, M., editors. Lexical and syntactic constructions and
the construction of meaning. John Benjamins; Amsterdam: 1997. p. 329-346.
Naigles L, Eisenberg A, Kako E, Highter M, McGraw N. Speaking of motion: Verb use in English and
Spanish. Language and Cognitive Processes. 1998; 13:521–549.
Oakes LM. The development of infants’ use of continuity cues in their perception of causality.
Developmental Psychology. 1994; 30:748–756.
Ogura, T.; Murase, Y. Communicative development inventory; Congress of Japanese Developmental
Psychology; Tokyo, Shimane University, Matsue. 1991;
Otsuka Y, Yamaguchi MK. Infants’ perception of illusory contours in static and moving figures.
Journal of Experimental Child Psychology. 2003; 86:244–251. [PubMed: 14559206]
Parish, J.; Pruden, SM.; Ma, W.; Hirsh-Pasek, K.; Golinkoff, RM. A world of relations: Relational
words. In: Malt, B.; Wolff, P., editors. Words and the mind: How words capture human
experience. Oxford University Press; New York, NY: 2010. p. 219-242.
Peterson MA, Gibson BS. Must figure-ground organization precede object recognition? An
assumption in peril. Psychological Science. 1994; 5:253–259.
Phillips AT, Wellman HM, Spelke ES. Infants’ ability to connect gaze and emotional expression to
intentional action. Cognition. 2002; 85:53–78. [PubMed: 12086713]
Poulin-Dubois D, Shultz TR. The infant’s concept of agency: The distinction between social and
nonsocial objects. The Journal of Genetic Psychology. 1990; 151:77–90. [PubMed: 2332761]
Pruden SM, Göksun T, Roseberry S, Hirsh-Pasek K, Golinkoff RM. Find your manners: How do
infants detect the invariant manner of motion in dynamic events? Child Development. in press.
Pruden, SM.; Hirsh-Pasek, K.; Maguire, MJ.; Meyer, MA. In: Brugos, A.; Micciulla, L.; Smith, CE.,
editors. Foundations of verb learning: Infants form categories of path and manner in motion
events; Proceedings of the 28th annual Boston University Conference on Language Development;
Somerville, MA: Cascadilla Press. 2004; p. 461-472.
Pulverman R, Golinkoff RM, Hirsh-Pasek K, Buresh JS. Infants discriminate paths and manners in
nonlinguistic dynamic events. Cognition. 2008; 108:825–830. [PubMed: 18599030]
Rakison DH, Poulin-Dubois D. The developmental origin of the animate–inanimate distinction.
Psychological Bulletin. 2001; 127:209–228. [PubMed: 11316011]
Robertson SS, Suci GJ. Event perception by children in the early stages of language production. Child
Development. 1980; 51:89–96. [PubMed: 7363753]
Sasson NJ, Turner-Brown LM, Holtzclaw TN, Lam KSL, Bodfish JW. Children with autism
demonstrate circumscribed attention during passive viewing of complex social and nonsocial
picture arrays. Autism Research. 2008; 1:31–42. [PubMed: 19360648]
Sharon T, Wynn K. Individuation of actions from continuous motion. Psychological Science. 1998;
9:357–362.
Shipley, TF.; Zacks, JM., editors. Understanding events: From perception to action. Oxford University
Press; New York: 2008.
Simons DJ, Levine DT. Failure to detect changes to people during a real-world interaction.
Psychonomic Bulletin and Review. 1998; 5:644–649.
Slobin, DI. Linguistic representations of motion events: What is signifier and what is signified?. In:
Maeder, C.; Fischer, O.; Herlofsky, W., editors. Iconicity inside out: Iconicity in language and
literature 4. John Benjamins; Amsterdam/Philadelphia: 2005. p. 307-322.
Sodian B, Schoeppner B, Metz U. Do infants apply the principle of rational action to human agents?
Infant Behavior and Development. 2004; 27:31–41.
Spelke ES, Born WS, Chu F. Perception of moving, sounding objects by four-month-old infants.
Perception. 1983; 12:719–732. [PubMed: 6678415]
Göksun et al. Page 27
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Talmy, L. Lexicalization patterns: Semantic structure in lexical forms. In: Shopen, T., editor.
Language typology and syntactic description. Cambridge University Press; New York, NY: 1985.
p. 57-149.
Talmy, L. Concept structuring systems. Vol. I. MIT Press; Cambridge, MA: 2000. Toward a cognitive
semantics; p. i-viii.p. 1-565.
Tomasello, M.; Merriman, WE., editors. Beyond names for things: Young children’s acquisition of
verbs. Lawrence Erlbaum Associates; Hillsdale, NJ: 1995.
Tsao FM, Liu HM, Kuhl PK. Speech perception in infancy predicts language development in the
second year of life: a longitudinal study. Child Development. 2004; 75:1067–1084. [PubMed:
15260865]
Tsujimura, N. An introduction to Japanese linguistics. Blackwell; NY: 2006.
Wagner L, Carey S. 12-month-old infants represent probable endings of motion events. Infancy. 2005;
7:73–83.
Wagner L, Lakusta L. Using language to navigate the infant mind. Perspectives on Psychological
Science. 2009; 4:177–184. [PubMed: 20161142]
Wang S, Baillargeon R, Paterson S. Detecting continuity violations in infancy: A new account and new
evidence from covering and tube events. Cognition. 2005; 95:129–173. [PubMed: 15694644]
Werker JF, Tees RC. Cross-language speech perception: Evidence for perceptual reorganization during
the first year of life. Infant Behavior and Development. 1984; 7:49–63.
Wever EG. Figure and ground in the visual perception of form. The American Journal of Psychology.
1927; 38:194–226.
Whorf, BL. Language, thought, and reality. MIT Press; Cambridge, MA: 1956.
Woodward AL. Infants selectively encode the goal object of an actor’s reach. Cognition. 1998; 69:1–
34. [PubMed: 9871370]
Woodward AL. Infants’ developing understanding of the link between looker and object.
Developmental Science. 2003; 6:297–311.
Wynn K. Infants’ individuation and enumeration of actions. Psychological Science. 1996; 7:164–169.
Göksun et al. Page 28
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Figure 1.
a. Figures (a girl, a boy, a man, and a woman) used in the figure discrimination study.
b. Grounds used in the ground discrimination study. The grounds in the top panel (railroad
track, street, road, and bridge) are encoded by the Japanese verb wataru ‘a flat barrier
dividing two points’ and grounds in the bottom panel (tennis court and grass) are coded by
the verb tooru ‘a continuous plane.’
Göksun et al. Page 29
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Figure 2.
The design and sample stimuli for the figure (top panel) and ground (bottom panel)
discrimination studies
Göksun et al. Page 30
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Figure 3.
Eight- and 11-month-old infants’ mean percentage of looking times in the test phase to
novel vs. familiar figures (top panel) and novel vs. familiar grounds (bottom panel). *p < .
05.
Göksun et al. Page 31
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Figure 4.
Eight- and 11-month-olds’ mean percentage of looking times to novel and familiar grounds
at test in within- and across-category comparisons (ps > .05).
Göksun et al. Page 32
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Figure 5.
Fourteen-month-olds’ mean percentage of looking times to novel and familiar grounds at
test in within- and across-category comparisons. *p < .05.
Göksun et al. Page 33
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Figure 6.
Fourteen-month-olds’ mean percentage of looking times to novel and familiar grounds at
test in the within- and across-category comparisons in the ground discrimination study with
grayscale. *p < .05.
Göksun et al. Page 34
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Figure 7.
Eight- and 11-month-olds’ mean percentage of looking times in the test phase with static
displays to figures (top panel) and grounds (bottom panel). *p < .05.
Göksun et al. Page 35
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Figure 8.
English- and Japanese-reared 14- and 19-month-olds’ mean percentage of looking times to
novel and familiar grounds at test in within-category (top panel) and across-category
(bottom panel) comparisons (*p < .05).
Göksun et al. Page 36
Cognition. Author manuscript; available in PMC 2012 November 1.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript