Sixteen-month-olds communication
can use language to update
their expectations about the visual world
Patricia A. Ganea a,⇑
, Allison Fitch b
, Paul L. Harris c
, Zsuzsa Kaldy b
aUniversity of Toronto, Toronto, Ontario M5S 3G3, Canada
bUniversity of Massachusetts Boston, Boston, MA 02125, USA
c Harvard University, Cambridge, MA 02138, USA
article info
Article history:
Available online xxxx
Keywords:
Mental representation
Updating
Absent reference
Language
Object location
Visual world
abstract
The capacity to use language to form new representations and to
revise existing knowledge is a crucial aspect of human cognition.
Here we examined whether infants can use language to adjust
their representation of a recently encoded scene. Using an eyetracking paradigm, we asked whether 16-month-old infants
(N = 26; mean age = 16;0 [months;days], range = 14;15–17;15)
can use language about an occluded event to inform their expectation about what the world will look like when the occluder is
removed. We compared looking time to outcome scenes that
matched the language input with looking time to those that did
not. Infants looked significantly longer at the event outcome when
the outcome did not match the language input, suggesting that
they generated an expectation of the outcome based on that input
alone. This effect was unrelated to infants’ vocabulary size. Thus,
using language to adjust expectations about the visual world is
present at an early developmental stage even when language skills
are rudimentary.
! 2016 Elsevier Inc. All rights reserved.
Introduction
Much of the information that we have about the world is based on testimony provided by other
people. Thus, on the basis of other people’s testimony, we can acquire knowledge about events that
http://dx.doi.org/10.1016/j.jecp.2015.12.005
0022-0965/! 2016 Elsevier Inc. All rights reserved.
⇑ Corresponding author.
E-mail address: [email protected] (P.A. Ganea).
Journal of Experimental Child Psychology xxx (2016) xxx–xxx
Contents lists available at ScienceDirect
Journal of Experimental Child
Psychology
journal homepage: www.elsevier.com/locate/jecp
Please cite this article in press as: Ganea, P. A., et al. Sixteen-month-olds can use language to update their expectations about the visual world. Journal of Experimental Child Psychology (2016), http://dx.doi.org/10.1016/j.
jecp.2015.12.005
we would otherwise know little or nothing about—for example, events in the distant past. In addition,
however, other people’s testimony can help us to update our knowledge of the current state of the
world (Harris, 2012). Indeed, we often rely on such testimony to provide us with information about
changes that have occurred with respect to people or objects that are already known to us. For example, via testimony, friends and family members often provide us with updates about their lives. They
can tell us about changes in their love lives, their jobs, their kitchen appliances, and so forth. By implication, updating based on verbal testimony has a broad scope.
Developmental research has provided a wealth of information on infants’ ability to update object
representations on the basis of visual information (e.g., Feigenson & Yamaguchi, 2009; Koechlin,
Dehaene, & Mehler, 1997; Uller, Carey, Huntley-Fenner, & Klatt, 1999; Wellman, Cross, & Bartsch,
1986; Wynn, 1992) such as when changes are being made to the number of objects in a scene or to
an object’s location in the scene. However, only recently have we begun to learn information about
children’s ability to update their knowledge of an object through language. Ganea and Harris (2010,
2013) showed that when toddlers aged 30 months were told that an object they had put in one container had been moved to a different container during their temporary absence, they searched for the
object in the new container, not where they had originally put it. By implication, using the testimony
provided by another person, they updated their own prior representation of the state of the world and
searched correctly.
However, in those same studies, younger toddlers—aged approximately 24 months—were much
less accurate in making such location updates. They often searched for the object on the basis of their
earlier firsthand observation of its whereabouts—an error that did not occur in a control condition
where they saw the object moved to a new location rather than learning about its movement via verbal testimony. Alternatively, they first searched in the original container and only subsequently went
on to search at the object’s new location. Thus, these younger infants were less likely than older
infants to give priority to what they had been told.
One possible interpretation of this age change is that younger infants have difficulty in updating
their representation of the world on the basis of verbal input. To the extent that a great deal of conversational input to young children is concerned with objects, events, and outcomes that are present
and potentially observable within the immediate situation (Morford & Goldin-Meadow, 1997), such a
restriction might not create any obvious cognitive difficulty. According to this hypothesis, we might
expect children’s updating ability to improve as their early language becomes increasingly displaced
from the here and now.
However, despite the above findings, updating may emerge considerably earlier in development.
There are several reasons for considering this alternative possibility. First, task demands may have
underestimated the ability of younger toddlers in the studies of Ganea and Harris (2010, 2013). They
may have understood what they were told and updated their representation of the object’s location
but failed to inhibit a prepotent response of searching in the place where they had last seen the object.
Under this interpretation, infants can engage in language-based updating but have difficulty in using
the updated representation of location to guide voluntary actions such as search. Similar gaps between
the representation of location and the execution of a response have been well established in studies of
infants’ ability to search for a hidden object (Diamond, 1985).
Second, by the beginning of the second year of life, infants are adept at incorporating testimony
into their expectations about new unseen referents. For example, infants are able to reason about
the quantities of unseen objects and nonobvious functions of objects through information provided
via language alone (Graham, Kilbreath, & Welder, 2004; Xu, Cote, & Baker, 2005).
Third, studies of infants’ ability to engage in updating with respect to an object’s properties—as
opposed to its location—have yielded encouraging results. For example, in one study infants left a
stuffed animal in a room and went next door. They were subsequently told that the stuffed animal
had been accidentally made wet. When they went back to retrieve it, 22-month-olds picked out a
wet version of the stuffed animal rather than a dry version (Ganea, Shutts, Spelke, & DeLoache,
2007). Furthermore, follow-up studies on property updating demonstrated that reducing task
demands by strengthening the working memory representation of the to-be-updated object and
increasing its familiarity promoted updating in even younger toddlers, with 19-month-olds succeeding in updating object properties (Galazka & Ganea, 2014).
2 P.A. Ganea et al. / Journal of Experimental Child Psychology xxx (2016) xxx–xxx
Please cite this article in press as: Ganea, P. A., et al. Sixteen-month-olds can use language to update their expectations about the visual world. Journal of Experimental Child Psychology (2016), http://dx.doi.org/10.1016/j.
jecp.2015.12.005
Taken together, these considerations suggest that even younger infants—provided that they can
understand a simple assertion about object movement—might update their expectations about the
location of an object based on what they are told. The current violation-of-expectation study tested
this prediction using a computer presentation. The violation-of expectation paradigm used here
was an adaptation of the blank-screen paradigm, which was originally designed to record eye movements as adult participants listen to a sentence when the screen is blank (Altmann & Kamide, 2004).
This paradigm allows one to measure patterns of eye movements when the visual scene is absent (e.g.,
whether infants look to locations of mentioned absent objects) and to measure whether looks to the
scene when visible are influenced by prior language. Accordingly, in this study we wanted to explore
infants’ looking patterns when the visual scene was absent at the time of the linguistic input and also
when the visual scene was available again.
Sixteen-month-old infants were first familiarized with two pairs of common animals (two dogs and
two cats) on a computer screen. Next, they were presented with a scene that included a target location—for example, a bed—and two different familiar animals, always one dog and one cat, located on
either side of the target location. Infants heard a speaker call their attention to each animal and to the
target location. Then children watched as a curtain went down to hide the scene and the speaker
described an invisible action of one of the two animals—for example, ‘‘Now the dog goes to the table.
The dog is on the table.” The curtain was then raised to reveal a scene that was either congruent with
the speaker’s testimony (i.e., the dog on the table and the cat still in its original position) or incongruent with her testimony (i.e., the cat on the table and the dog still in its original position). If infants successfully updated their representation of the scene based on what they were told while the curtain
was lowered, they should be surprised by the incongruent scene but not by the congruent scene.
We tested infants’ expectations by comparing their looking time patterns to the incongruent versus
congruent scene.
Method
Participants
The participants were 41 healthy, full-term, 16-month-old infants (age range = 14;15–17;15
[months;days]). Of these participants, 11 did not complete the required two blocks of trials as a result
of fussiness and were excluded from further analysis. In addition, 3 infants were excluded due to
experimenter error, and 1 infant was excluded due to equipment failure. The final sample consisted
of 26 infants (12 female and 14 male; average age = 16;0, SD = 27.2 days). Of this final sample, 16
infants were Caucasian (1 of them Hispanic), 3 were African American, 2 were biracial, and 5 families
chose not to report their race and ethnicity. Parents of participants were recruited from the greater
Boston area in the northeastern United States based on data from state birth records. They received
$20 and a small gift for participation. All participants were from dominantly (>75%) Englishspeaking households, with receptive vocabularies ranging from 11 to 350 words (M = 153.54,
SD = 75.51) and productive vocabularies ranging from 0 to 66 words (M = 26.83, SD = 20.81) as measured by the MacArthur–Bates Communication Development Inventory (MCDI). Statistics on MCDI
items pertinent to the study’s design are displayed in Table 1.
Apparatus
We used a Tobii T120 eye-tracker running Tobii Studio 3.0 software (Tobii Technology, Stockholm,
Sweden) sampling at 60 Hz to measure eye movement patterns. Eye position was determined by averaging data from both eyes. When data could be acquired from only one eye (e.g., when the infant’s face
was momentarily not centered on the screen), they were not used. Fixations were defined using Tobii
Studio’s standard built-in fixation filter (Tobii Fixation Filter), which automatically interpolates segments of missing data that are shorter than 100 ms. Participants sat on their parents’ laps approximately 60 cm away from the built-in 17-inch monitor in a dimly lit testing room. Parents’ eyes
were shielded by a visor, by modified sunglasses with occluded lenses, or by parents closing their eyes.
P.A. Ganea et al. / Journal of Experimental Child Psychology xxx (2016) xxx–xxx 3
Please cite this article in press as: Ganea, P. A., et al. Sixteen-month-olds can use language to update their expectations about the visual world. Journal of Experimental Child Psychology (2016), http://dx.doi.org/10.1016/j.
jecp.2015.12.005
Stimuli and procedure
We employed a within-participants design with two blocks of test trials: Congruent versus Incongruent trials. Each block consisted of 4 familiarization trials followed by 6 test trials. Participants saw
both blocks of trials, with a short break in between for feeding, changing, or a brief walk down the
hallway. (Breaks varied in length based on the individual infants’ needs but were typically between
5 and 10 min.) The order of the two blocks was counterbalanced across participants. The blocked
design was chosen to maximize the likelihood of being able to observe a difference between the
two conditions. At the beginning of each block, each participant was calibrated using a standard 5-
point calibration procedure. Each block took approximately 2.5 min. Agents (cats and dogs) and locations (table and bed) were quasi-randomly drawn from the same two sets of cats and dogs and locations (table and bed) in each trial.
In familiarization trials, a curtain was raised to reveal either two cats or two dogs (counterbalanced
in order of presentation) for 4 s. Right after the curtain went up, a female voice (prerecorded by a
native English speaker in child-directed speech) said, ‘‘Look at the cats!” or ‘‘Look at the dogs!” Test
trials consisted of three phases: an initial exposure, an updating period, and an outcome period. In
the initial exposure, a curtain was raised to reveal an object in the middle of the stage (either a bed
or a table) and a cat and a dog on each side of the object (side counterbalanced and order quasirandomized; see Fig. 1). The female voice said, ‘‘Look at the cat! Look at the table/bed! Look at the
dog!” This initial presentation of the scene, along with the recording, lasted for 8 s. Then a curtain lowered, occluding the scene, displaying only three identical triangles as placeholders for the positions of
the animals/object. These triangles, which remained static throughout the test phase, served as anchor
points for the referents and allowed participants to access each referent’s spatial location during
speech processing, as has been demonstrated with adults (Altmann & Kamide, 2004; Ferreira, Apel,
& Henderson, 2008). Following the descent of the curtain, the female voice said, ‘‘Now the dog goes
to the table. The dog is on the table.” The curtain remained in the lowered position while the testimony was given for 8 s (henceforth the updating period). Then the curtain was raised again to reveal
either a scene that was congruent with the statement during the updating period (the dog was on the
table, whereas the cat remained in its original position; Congruent condition) or incongruent with the
statement (the cat was on the table, whereas the dog remained in its original position; Incongruent
condition). This outcome scene was presented for 7 s (henceforth the outcome period), and then
the curtain was lowered and the trial ended. A sample video of a participant during the Incongruent
condition is provided in the online Supplementary material to illustrate the procedure.
Data analysis
To analyze gaze patterns, we defined two rectangular (327 by 268 pixels) areas of interest (AOIs)
around the two animals (the cat and the dog; see Fig. 1) during the outcome period and defined one
rectangular (327 by 268 pixels) AOI around each triangle placeholder during the updating period. AOIs
were defined a priori to contain the entire animal during the outcome period plus some additional
space around them to allow for small calibration errors. The AOIs for the triangle placeholders during
the updating period were larger, such that the adjacent AOIs filled the space between them without
any gaps. Total looking time to each AOI (defined by the cumulative duration of visits within an
AOI) as a proportion of looking to the entire screen was analyzed for both the updating and outcome
Table 1
Summary statistics of MCDI results for names of agents and objects used in experiment.
Item Understands Understands and says Does not understand
Dog 10 12 2
Cat 14 7 3
Table 0 10 14
Bed 18 2 4
Note. Values in table are numbers of participants.
4 P.A. Ganea et al. / Journal of Experimental Child Psychology xxx (2016) xxx–xxx
Please cite this article in press as: Ganea, P. A., et al. Sixteen-month-olds can use language to update their expectations about the visual world. Journal of Experimental Child Psychology (2016), http://dx.doi.org/10.1016/j.
jecp.2015.12.005
periods (see Results). Trials in which a 1-s cumulative looking time threshold was not met were
excluded from analysis (42 of 312 trials); participants who did not meet this threshold in at least 3
of 6 trials in both blocks were likewise excluded (11 participants, listed under excluded participants).
The 1-s threshold was chosen as a conservative estimate of the minimum length of time needed for
infants to process each object and its corresponding location. No other data filtering was applied.
Results
For each participant, a proportion of total looking time during the outcome period of each condition
was calculated by dividing the total looking time to each AOI (e.g., the animal on top of the table/bed)
by the total looking time to the whole screen and then averaging across the 6 trials. Proportion of looking time was calculated only for valid trials (at least 1 s of cumulative looking to the screen during the
outcome period). All participants completed at least 3 of the 6 trials per block (M = 5.19 trials in the
Congruent condition and M = 5.11 trials in the Incongruent condition), and the number of trials completed did not significantly differ between conditions, t(25) = 0.278, p = .78, d = 0.05. Mean looking
times during the outcome and updating periods are presented in Table 2.
Outcome period
To identify a violation-of-expectation (VoE) effect, we conducted a 2 (Condition) ! 2 (Order of
Block Presentation) mixed design analysis of variance (ANOVA) on the proportion of looking time to
the outcome AOI (the animal in the outcome location, i.e., on top of the table/bed). Data are shown
Look at the cat!
Look at the table!
Look at the dog!
Now the dog goes to the
table. The dog is on the table.
unmentioned mentioned center
on top
on side
Fig. 1. Schematic of a test trial (Incongruent condition). Red rectangles and tags above them indicate AOIs; they were not visible
to participants. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this
article.)
P.A. Ganea et al. / Journal of Experimental Child Psychology xxx (2016) xxx–xxx 5
Please cite this article in press as: Ganea, P. A., et al. Sixteen-month-olds can use language to update their expectations about the visual world. Journal of Experimental Child Psychology (2016), http://dx.doi.org/10.1016/j.
jecp.2015.12.005
in Fig. 2. Results indicated a main effect of condition, F(1, 24) = 44.16, p < .001, gp
2 = .648, where participants showed a significantly longer looking time in the Incongruent condition (M = .68, SD = .12) than
in the Congruent condition (M = .46, SD = .11). There was no significant main effect of order, F(1, 24)
= 2.02, p = .17, gp
2 = .078, or interaction between condition and order, F(1, 24) = 0.02, p = .89, gp
2 = .001.
Proportion of looking time to the outcome AOI (the animal that was on top of the target location) was
also significantly above chance in the Incongruent condition, t(25) = 7.48, p < .001, d = 1.41, and at
chance in the Congruent condition, t(25) = "1.56, p = .13, d = 0.32. Finally, participants were categorized into those who demonstrated a VoE effect (i.e., those who looked longer in the Incongruent condition than in the Congruent condition) and those who did not. Of the 26 participants, 23
demonstrated a VoE effect. A chi-square goodness-of-fit test determined that this is significantly different from chance, v2
(1) = 15.385, p < .001.
To explore the robustness of the VoE effect observed above, we repeated the 2 (Condition) ! 2
(Order of Block Presentation) mixed design ANOVA on proportion of looking time to the outcome
AOI on just the first trial in each block.1 The effect was present from the first trial; the main effect of
condition was significant, F(1, 24) = 5.143, p = .033, gp
2 = .176; that is, participants looked longer at the
outcome AOI in the Incongruent condition (M = .554, SD = .17) than in the Congruent condition
(M = .446, SD = .20). No significant main effect of order was found, F(1, 24) = 0.53, p = .473, gp
2 = .022,
and condition and order did not interact, F(1, 24) = 0.11, ns, gp
2 = .005.
Of the 24 participants with completed MCDIs, the overwhelming majority reportedly understood
the words ‘‘dog”, ‘‘cat”, and ‘‘bed”, but only 10 were reported to understand the word ‘‘table” (see
Table 1). To test whether the VoE effect was dependent on the destination of the location change
(i.e., only those who understood ‘‘table” showed a VoE for trials where a table was presented), trials
in which the table was the destination were analyzed separately. We performed a 2 (Condition) ! 2
(Understands ‘‘Table” vs. Does Not Understand ‘‘Table”) mixed design ANOVA on proportion of looking
time for these trials only. The trend from the previous analyses was present (albeit with a smaller
effect) in this subset of our data; the main effect of condition was marginally significant, F(1, 22)
= 4.125, p = .055, gp
2 = .158, in that the proportion of looking to the outcome AOI was larger in the
Incongruent condition (M = .56, SD = .13) than in the Congruent condition (M = .48, SD = .13). There
was no main effect of understanding the word ‘‘table”, F(1, 22) = 0.319, p = .578, gp
2 = .014, and this factor did not interact with condition, F(1, 22) = 1.352, p = .257, gp
2 = .058.
Updating period
To examine whether the differences in looking time during the outcome period were reflected in
gaze patterns when the predictive expectation may have been generated, we also examined looking
time differences during the updating (screen down) period. We hypothesized that if predictive expectations occurred, participants would display longer looking to the central AOI as well as the AOI of the
mentioned animal when those referents were stated in the testimony, reflecting adult-like speech processing. Fig. 3 plots a time course of the proportion of looks to each AOI by condition during this
period.
We conducted a 2 (Condition) ! 2 (Order of Block Presentation) ! 3 (AOI) mixed design ANOVA on
the proportion of total looking. Data are shown in Fig. 4. Results indicated a main effect of AOI, F(2, 48)
= 35.59, p < .001, gp
2 = .597. Pairwise comparisons (Bonferroni adjusted for multiple comparisons)
Table 2
Mean looking times (in seconds) to test trials by phase and condition.
Condition Encoding phase Testimony phase Outcome phase
Congruent 5.89 (1.75) 5.1 (1.85) 5.79 (1.10)
Incongruent 5.03 (1.67) 5.19 (1.35) 5.63 (1.57)
Note. Standard deviations are in parentheses.
1 We also analyzed whether looking times changed over time within a block and found no significant trends.
6 P.A. Ganea et al. / Journal of Experimental Child Psychology xxx (2016) xxx–xxx
Please cite this article in press as: Ganea, P. A., et al. Sixteen-month-olds can use language to update their expectations about the visual world. Journal of Experimental Child Psychology (2016), http://dx.doi.org/10.1016/j.
jecp.2015.12.005
demonstrated that this effect was driven by looking to the center AOI (M = .421, SD = .12), which was
significantly longer than looking to both the mentioned AOI (M = .202, SD = .076, p < .001) and the
unmentioned AOI (M = .188, SD = .081, p < .001). There were no significant differences between the
looking times to the mentioned and unmentioned AOIs (p > .99).
This analysis also yielded a significant main effect of condition, F(1, 24) = 9.26, p = .006, gp
2 = .278, in
that participants spent a slightly larger portion of their total looking time within the three AOIs in the
Congruent condition (for any given AOI: M = .284, SD = .035) than in the Incongruent condition (for
any given AOI: M = .257, SD = .035). Arguably, infants in the Incongruent condition became more
aroused over time because they were presented with a conflict between sentences and outcomes
on 6 successive test trials. Such a repeated violation of expectation may have led them to look around
randomly during the updating (screen down) period. There was no significant main effect of order, F(1,
24) = 0.007, p = .94, gp
2 < .001, and no interactions.
Although participants looked overall at the center AOI more than at the other two AOIs during the
updating period, we investigated the possibility that infants’ gaze behavior may have changed right
around the time when they heard the agent being mentioned during the testimony. To do so, we compared the proportion of looking to the mentioned agent during three 500-ms time windows—premention (500 ms before the first mention), post-first mention (500 ms after the first mention), and
post-second mention (500 ms after the second mention)—averaged across trials within a block. We
conducted a 2 (Condition) ! 3 (Time Window) repeated measures ANOVA on the proportion of total
looking to the mentioned AOI. Results indicated that there was no significant main effect of condition,
F(1, 19) = 0.225, p = .64, gp
2 = .012, or time window, F(2, 38) = 0.926, p = .41, gp
2 = .046. There was also no
interaction between condition and time window, F(2, 38) = 0.942, p = .39, gp
2 = .047. In the Incongruent
condition, the average proportion of time spent on the mentioned agent was .257 (.215) during the
pre-mention period, .286 (.314) during the first post-mention period, and .269 (.255) during the second post-mention period. In the Congruent condition, these values were .289 (.256), .262 (.258), and
.184 (.168), respectively.
Finally, we went on to explore possible developmental contributions to the VoE effect observed
during the outcome period. To do this, we examined relationships between the magnitude of VoE
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Congruent Incongruent
Proportion of looking time to outcome AOI
Condition
Congruent First
Incongruent First
Fig. 2. Proportions of looking time to outcome AOI (animal in the outcome location) during the outcome period as a function of
condition and block order. Error bars display standard errors of the means. ***p < 0.001.
P.A. Ganea et al. / Journal of Experimental Child Psychology xxx (2016) xxx–xxx 7
Please cite this article in press as: Ganea, P. A., et al. Sixteen-month-olds can use language to update their expectations about the visual world. Journal of Experimental Child Psychology (2016), http://dx.doi.org/10.1016/j.
jecp.2015.12.005
(the difference in proportion of looking to the event AOI between the Incongruent and Congruent conditions) and age (in days) as well as receptive vocabulary size on the MCDI.2 The correlations between
VoE effect magnitude and age (Pearson’s r = ".06, p = .78) and receptive vocabulary (r = ".27, p = .21)
were not significant.
Discussion
The capacity to use language to acquire knowledge is a fundamental aspect of human cognition.
Here we examined an important aspect of that knowledge acquisition process, namely the ability to
update one’s representation of the world on the basis of verbal testimony rather than direct observation. Thus, we asked whether infants can update their representation of a recently encoded visual
scene based on a verbal description of a change in that scene. The visual scene contained two referent
animals with an inanimate object located between them. When the visual scene was occluded, infants
Fig. 3. Proportion of looks to each AOI over time during the updating period in the Congruent condition (A) and the Incongruent
condition (B). Testimony and 500-ms time windows (pre-mention, post-first mention, and post-second mention) are overlaid.
2 Two children were not included in the MCDI analyses. The parent of one of these children chose not to fill out the
questionnaire; another parent misunderstood the instructions, yielding an extreme score.
8 P.A. Ganea et al. / Journal of Experimental Child Psychology xxx (2016) xxx–xxx
Please cite this article in press as: Ganea, P. A., et al. Sixteen-month-olds can use language to update their expectations about the visual world. Journal of Experimental Child Psychology (2016), http://dx.doi.org/10.1016/j.
jecp.2015.12.005
were told about the change in the scene, specifically a change in the location of one of the referent
animals. When the visual scene was accessible again, infants looked longer at the event outcome
(the animal on top of the inanimate object), relative to the other animal on the screen, when this outcome did not match the state of affairs described by the prior language. When the event outcome was
congruent with the prior language, looking at the outcome event was at chance level. These results
indicate that infants used the language input to generate an expectation of the described outcome
and that they were surprised when the outcome did not match that input. This effect was not related
to infants’ age or vocabulary size.
These findings add to an increasing body of research showing that from the earliest stages of language development, children have the capacity to treat language as a source of information about
changes in the world that they have not witnessed firsthand. They can use verbal testimony to update
their representation of the world.
During the first half of their second year of life, infants understand that words can refer to absent
objects (Ganea, 2005; Ganea & Saylor, 2013a; Osina, Saylor, & Ganea, 2013, 2014; Saylor & Baldwin,
2004) and that words can refer not only to particular objects but also to general kinds (Gelman,
Hollander, Star, & Heyman, 2000; Graham, Booth, & Waxman, 2012; Waxman & Gelman, 2010). By
18 months of age, infants can also use language about a change in the properties or location of an
absent object to subsequently re-identify or search for the object (Galazka & Ganea, 2014; Ganea &
Harris, 2013). Nevertheless, if an object was originally encoded in a specific location (in the drawer
rather than elsewhere in the room), it is not until approximately 30 months of age that infants show
evidence of updating by searching systematically in the new location rather than in the old location
(Ganea & Harris, 2010, 2013).
The current research shows that even at 16 months of age, infants can use language to form expectations about an object’s current location in a visual scene. Infants’ looking times to the event outcome
were significantly longer when the visual scene did not match the language they received about the
scene in its absence than when it did. Infants’ responses at test were not simply due to having
observed any change in the new visual scene relative to the already encoded visual scene because
the visual scene presented at test was changed from the familiarization phase in both the Congruent
and Incongruent conditions. More specifically, the scene always showed one of the two animals having moved from the side onto the object at the center. Thus, the language received in the absence of
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Mentioned Center Unmentioned
Proportion of looking time
Congruent
Incongruent
Fig. 4. Proportion of looking to each AOI during the testimony period as a function of condition. Error bars display standard
errors of the means. ***p < 0.001.
P.A. Ganea et al. / Journal of Experimental Child Psychology xxx (2016) xxx–xxx 9
Please cite this article in press as: Ganea, P. A., et al. Sixteen-month-olds can use language to update their expectations about the visual world. Journal of Experimental Child Psychology (2016), http://dx.doi.org/10.1016/j.
jecp.2015.12.005
the scene influenced infants’ expectation of the outcome in the scene (with one of the referent animals
on top of the object rather than in its original location).
The language used in this study predicted a relation between two of the referent objects in the
scene (‘‘Now the dog goes to the table. The dog is on the table”) and, as a result, a change in the location of one of the referents (the dog). As children heard the sentence, did they mentally represent the
dog’s prior location in the previously visible scene and then update their representation of the dog’s
location based on the language input? Alternatively, did infants create a new representation of the
scene (of a dog on the table) at the time they heard the language input with no updating of the prior
representation involved? In both cases, infants would have been surprised when the visual input did
not match the language they heard when the occluder was lowered.
One reason to prefer the first alternative is that studies with adults using the blank screen paradigm have shown that when a scene is absent, adults look where a mentioned object had been visible
and update the representation of the scene as the language is being processed (Altmann & Kamide,
2004). The data during the updating period in our study (when the scene was no longer visible) show
that infants looked at the location where the table or bed had been visible significantly longer than at
the prior locations of the animals, but the proportion of looks to the mentioned agent was not higher
than that to the unmentioned agent overall (Fig. 4), and our time course analysis demonstrated that it
did not significantly increase around the time of the mentions (Fig. 3). This implies that infants did not
process the testimony in a fully adult-like way; that is, unlike adults, they did not look at the mentioned animal’s placeholder while the scene was occluded. Yet this is not surprising in light of the evidence that toddlers do not display adult-like gaze behavior during sentence processing until they are
approximately 8 months older and have a sizable (>225 words) productive vocabulary (Mani &
Huettig, 2012). The largest productive vocabulary among the infants in our study was a mere 66
words.
Studies on infants’ understanding of references to absent objects show that, beginning at
12 months of age, infants show comprehension of references to absent objects by looking, pointing,
or searching at the object’s likely location on hearing its name mentioned (Ganea, 2005; Ganea &
Saylor, 2013b; Osina et al., 2013, 2014; Saylor & Baldwin, 2004). Infants are less likely to engage in
these communicative behaviors toward absent referents if they hear an unrelated name (Ganea,
2005). These results make it unlikely that the infants in the current study interpreted the sentence
afresh rather than as referring to the visual scene that they had seen only a few seconds prior to
the language input. Nevertheless, further research with infants is needed to demonstrate whether
the incremental language processing skills shown in adults emerge at the earliest stages of language
development. Even though children may activate a representation of the absent referent (for a review
of the evidence, see Ganea & Saylor, 2013a), it is not clear that at this age the timing of this activation
happens incrementally as the language is being processed.
Another important question regarding infants’ processing of absent reference concerns the nature
of the information extracted from the language that infants heard when the scene was not visible.
Research with older infants shows that by 24 months of age infants can use semantic information that
is language derived (e.g., whether the verb is transitive or intransitive) to determine the specific type
of event that a sentence describes (Arunachalam & Waxman, 2010). The language we used here specified the spatial relation between the animal and its destination (e.g., that the dog was on the table).
However, it was not possible to determine exactly what information was extracted from this language
and, therefore, how specific infants’ representation of the relation was.
As a baseline, it is clear that infants extracted the agent of the change (either cat or dog) from the
testimony because that formed the critical distinction between congruent and incongruent testimony.
However, we might expect the same pattern of results if the agent was all that the participants had
extracted from the testimony. In other words, if infants extracted only the dog from the sentence, they
may have expected any salient change involving the dog to be a possible outcome once the occluder
was lifted; this would yield longer looking when the location change involved the cat instead. Likewise, we do not know whether children would have been surprised, and thus looked longer, if the
scene outcome showed the dog ‘‘under” the table rather than ‘‘on” the table. Additional experiments
are already under way in our lab varying the amount of information participants are able to extract
from the testimony as well as the degree to which the outcome scene is congruent with the testimony.
10 P.A. Ganea et al. / Journal of Experimental Child Psychology xxx (2016) xxx–xxx
Please cite this article in press as: Ganea, P. A., et al. Sixteen-month-olds can use language to update their expectations about the visual world. Journal of Experimental Child Psychology (2016), http://dx.doi.org/10.1016/j.
jecp.2015.12.005
These will help to determine whether infants’ semantic knowledge and processing of semantic links
between words in a sentence constrains their representation of the outcome event.
Our findings, together with those of others (Gliga & Csibra, 2009; Graham et al., 2004;
Vouloumanos, Onishi, & Pogue, 2012; Xu et al., 2005), show that from the early stages of language
development infants take language to provide relevant information about entities in the world even
when those entities are not visible. More specifically, we have shown here that infants as young as
16 months have the capacity to update their expectations about the location of an object based on
information they hear in the absence of the object. This finding suggests that older children’s difficulty
in relying on language to search for an object in a new location in prior updating studies (Ganea &
Harris, 2010, 2013) may be due to difficulty in using an updated representation to guide their search
for the object, which is related to working memory limitations shown to play a role in children’s ability to update (Ozdemir, Gallant, & Ganea, 2015). It is also possible that the level of conflict here
between the old location and the new location of the target animal was not as strong as in previous
research (Ganea & Harris, 2010, 2013) given that the initial location of the target animal in the current
paradigm was not highlighted in any special way. Further research is needed to probe the sophistication of children’s ability to process linguistic information in the absence of a visual scene and to
update their mental representation of such scenes based on new input.
Acknowledgments
We thank Marisa Biondi, Annalisa Groth, and other members of the UMass Boston Baby Lab for
their help with data collection. The research presented here and the preparation of this manuscript
were supported by a Seed Grant from the Simons Foundation under the auspices of the Simons Center
for the Social Brain at MIT (Grant 319294) to Zsuzsa Kaldy and an Insight Grant from the Social
Sciences and Humanities Council in Canada to Patricia Ganea.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in the online version, at http://dx.doi.
org/10.1016/j.jecp.2015.12.005.
References
Altmann, G. T. M., & Kamide, Y. (2004). Now you see it, now you don’t: Mediating the mapping between language and the visual
world. In J. M. Henderson & F. Ferreira (Eds.), The interface of language, vision, and action: Eye movements and the visual world
(pp. 347–386). New York: Psychology Press.
Arunachalam, S., & Waxman, S. R. (2010). Meaning from syntax: Evidence from 2-year-olds. Cognition, 114, 442–446.
Diamond, A. (1985). Development of the ability to use recall to guide action, as indicated by infants’ performance on A-not-B.
Child Development, 56, 868–883.
Feigenson, L., & Yamaguchi, M. (2009). Limits on infants’ ability to dynamically update object representations. Infancy, 14,
244–262.
Ferreira, F., Apel, J., & Henderson, J. M. (2008). Taking a new look at looking at nothing. Trends in Cognitive Science, 12, 405–410.
Galazka, M. A., & Ganea, P. A. (2014). The role of representational strength in verbal updating: Evidence from 19- and 24-montholds. Journal of Experimental Child Psychology, 121, 156–168.
Ganea, P. A. (2005). Contextual factors affect absent reference comprehension in 14-month-olds. Child Development, 76,
989–998.
Ganea, P. A., & Harris, P. L. (2010). Not doing what you are told: Early perseverative errors in updating mental representations
via language. Child Development, 81, 457–463.
Ganea, P. A., & Harris, P. L. (2013). Early limits on the verbal updating of an object’s location. Journal of Experimental Child
Psychology, 114, 89–101.
Ganea, P. A., & Saylor, M. M. (2013a). Representational constraints on language development: Thinking and learning about
absent things. Child Development Perspectives, 7, 227–231.
Ganea, P. A., & Saylor, M. M. (2013b). Talking about the near and dear: Infants’ comprehension of displaced speech.
Developmental Psychology, 49, 1299–1307.
Ganea, P. A., Shutts, K., Spelke, E. S., & DeLoache, J. S. (2007). Thinking of things unseen: Infants’ use of language to update
mental representations. Psychological Science, 18, 734–739.
Gelman, S. A., Hollander, M., Star, J., & Heyman, G. D. (2000). The role of language in the construction of kinds. In D. Medin (Ed.).
Psychology of learning and motivation (Vol. 39, pp. 201–263). San Diego: Academic Press.
Gliga, T., & Csibra, G. (2009). One-year-old infants appreciate the referential nature of deictic gestures and words. Psychological
Science, 20, 347–353.
P.A. Ganea et al. / Journal of Experimental Child Psychology xxx (2016) xxx–xxx 11
Please cite this article in press as: Ganea, P. A., et al. Sixteen-month-olds can use language to update their expectations about the visual world. Journal of Experimental Child Psychology (2016), http://dx.doi.org/10.1016/j.
jecp.2015.12.005
Graham, S. A., Booth, A. E., & Waxman, S. R. (2012). Words are not features of objects: Only consistently applied nouns guide 4-
year-olds? Inferences about object categories. Language Learning and Development, 8, 1–11.
Graham, S. A., Kilbreath, C. S., & Welder, A. N. (2004). Thirteen-month-olds rely on shared labels and shape similarity for
inductive inferences. Child Development, 75, 409–427.
Harris, P. L. (2012). Trusting what you’re told: How children learn from others. Cambridge, MA: Belknap Press/Harvard University
Press.
Koechlin, E., Dehaene, S., & Mehler, J. (1997). Numerical transformations in five-month-old infants. Mathematical Cognition, 3,
89–104.
Mani, N., & Huettig, F. (2012). Prediction during language processing is a piece of cake—But only for skilled producers. Journal of
Experimental Psychology: Human Perception and Performance, 38, 843–847.
Morford, J. P., & Goldin-Meadow, S. (1997). From here and now to there and then: The development of displaced reference in
homesign and English. Child Development, 68, 420–435.
Osina, M., Saylor, M. M., & Ganea, P. A. (2013). When familiar is not better: 12-month-olds respond to talk about absent objects.
Developmental Psychology, 49, 138–145.
Osina, M. A., Saylor, M. M., & Ganea, P. A. (2014). Object locations, identity, and absent reference understanding at 12 months.
Infancy, 19, 65–81.
Ozdemir, B., Gallant, H., & Ganea, P. A. (2015). Do working memory and inhibition explain toddlers’ ability to verbally update
mental representations? In: Poster presented at the 2015 biennial meeting of the Society for Research in Child
Development, Philadelphia, PA.
Saylor, M. M., & Baldwin, D. A. (2004). Discussing those not present: Comprehension of references to absent caregivers. Journal
of Child Language, 31, 537–560.
Uller, C., Carey, S., Huntley-Fenner, G., & Klatt, L. (1999). What representations might underlie infant numerical knowledge.
Cognitive Development, 14, 1–36.
Vouloumanos, A., Onishi, K. H., & Pogue, A. (2012). Twelve-month-old infants recognize that speech can communicate
unobservable intentions. Proceedings of the National Academy of Sciences of the United States of America, 109, 12933–12937.
Waxman, S. R., & Gelman, S. A. (2010). Different kinds of concepts and different kinds of words: What words do for human
cognition. In D. Mareschal, P. Quinn, & S. Lea (Eds.), The making of human concepts (pp. 101–130). Oxford, UK: Oxford
University Press.
Wellman, H., Cross, D., & Bartsch, K. (1986). Infant search and object permanence: A meta-analysis of the A-not-B error.
Monographs of the Society for Research in Child Development, 51 [3, Serial No. 214].
Wynn, K. (1992). Addition and subtraction by human infants. Nature, 358, 749–750.
Xu, F., Cote, M., & Baker, A. (2005). Labeling guides object individuation in 12-month-old infants. Psychological Science, 16,
372–377.