Detailed Visual Memory Capacity Is Present Early - 麻省理工学院人工智能研究专业

Detailed Visual Memory Capacity Is Present Early
in Childhood

1
Katrina Ferrara

1
, Sarah Furlong

, Soojin Park

1,2

, and Barbara Landau

1Department of Cognitive Science, Johns Hopkins University

2心理学系, Yonsei University

关键词:
memory ﬁdelity

visual memory, cognitive development, memory capacity, 目的

认出,

开放访问

杂志

抽象的

Previous studies have shown that adults are able to remember more than 1,000 images with
great detail. 然而, little is known about the development of this visual capacity, nor its
presence early in life. This study tests the level of detail of young children’s memory for a
large number of items, adapting the method of Brady, Konkle, 阿尔瓦雷斯, and Oliva (2008).
Four- and six-year-old children were shown more than 100 images of everyday objects. 他们
were then tested for recognition of familiar items in a binary decision task. The identity of the
foil test item was manipulated in three conditions (类别, Exemplar, and State). 孩子们
demonstrated high accuracy across all conditions, remembering not only the basic-level
类别 (类别), but also unique details (Exemplar), and information about position and
arrangement of parts (状态). These ﬁndings demonstrate that children spontaneously encode
a high degree of visual detail. Early in life, visual memory exhibits high ﬁdelity and extends
over a large set of items.

介绍

The capacity to form and retrieve visual representations of objects is a central aspect of hu-
man cognition. Preserving detailed aspects of these object representations holds functional
importance for everyday activities: We may want to remember which kitchen implement had
been washed (a spoon or a spatula?), which particular category member was used in a recipe
(serrated knife or a smooth blade?), or in what particular state a given individual item was left
(toaster oven open or closed?). In support of these commonplace tasks, studies have shown
that the information capacity of visual memory in adults is surprisingly detailed as well as
impressively large (布雷迪, Konkle, 阿尔瓦雷斯, & Oliva, 2008; Konkle, 布雷迪, 阿尔瓦雷斯, & Oliva,
2010).

Is this capacity the outcome of adults’ massive exposure to objects and their regular
engagement in tasks that require object discrimination based on subtle differences? Or does
it reﬂect a signature characteristic of the human visual memory system, present even without
lifelong practice? 如果是这样, this capacity should be present in children, despite their more limited
need to remember detailed differences among objects. Little work has been done to investigate
the overall capacity and level of detail inherent in children’s visual memory, and whether the

Katrina Ferrara is now at the Center for Brain Plasticity and Recovery, 神经内科, Georgetown
大学, and Sarah Furlong is in the Clinical and Cognitive Psychology Graduate Program, University of North
Carolina at Chapel Hill. Park and Landau contributed equally.

引文: Ferrara, K., Furlong, S。, 公园,
S。, & Landau, 乙. (2017). Detailed Visual
Memory Capacity Is Present Early in
童年. 开放的心态: Discoveries
in Cognitive Science, 2(1), 14–25.
https://doi.org/10.1162/opmi_a_00014

DOI:
https://doi.org/10.1162/opmi_a_00014

补充材料:
https://doi.org/10.1162/opmi_a_00014

已收到: 22 一月 2017
公认: 24 八月 2017

利益争夺: 作者
declare that they have no conﬂicts of
interest with respect to their authorship
or the publication of this article.

通讯作者:
Katrina Ferrara
ferrara@cogsci.jhu.edu

版权: © 2017
麻省理工学院
在知识共享下发布
归因 4.0 国际的
(抄送 4.0) 执照

麻省理工学院出版社

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
哦
p
米

我
/

我

A
r
t
我
C
e
–
p
d

F
/

2
1
1
4
1
8
6
8
3
0
2
哦
p
米
_
A
_
0
0
0
1
4
p
d

我

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Visual Memory Capacity in Childhood Ferrara et al.

capacity itself undergoes developmental change. These questions are explored in the present
research by testing the extent of children’s visual memory for images of objects.

Existing research shows that adults have an impressive ability to remember previously
viewed images. A seminal study in the 1970s found that after viewing 10,000 scenes for only
5 seconds each, participants were 83% correct in identifying which of two images they had
见过 (Standing, 1973). Further work has shown that, even after several hours, adults can
remember details of particular objects they saw, suggesting that their memory stores more
than simply the gist of the image (沃尔夫, 1998). 例如, adults are able to distinguish
between a target object and similar distractors that belong to the same basic-level category
(2008)
(Castelhano & Henderson, 2005; 霍林沃斯, 2004). 最近, Brady et al.
exposed adults to a large number of objects, and then tested their memory using a forced-
choice task in which the identity of the foil was carefully manipulated. Participants ﬁrst viewed
images of 2,500 物体. At test, they were shown two images, only one of which they had
以前见过. The target image was paired with either an object from a different category (例如,
a sofa vs. a banana), an object from the same basic-level category (例如, a blue train vs. A
green train), or the same object in a different state (例如, an upright bucket vs. the same bucket
on its side). Across all three conditions, performance was quite high (92%, 88%, 和 87%,
分别). These ﬁndings show that adults can store thousands of object representations
(Brady et al., 2008; Konkle et al., 2010).

In contrast to the breadth of research in the adult literature, few studies have probed
the nature of visual memory capacity in childhood. Some have used visual recognition tests
to assess children’s memory (例如, Luciana & 纳尔逊, 1998; Rose, 费尔德曼, Futterweit, &
Jankowski, 1997), but these studies have typically focused on the ability to identify a famil-
iar target among a small set of alternatives (例如, four items). Standardized measures include
assessments of visual memory, but these are not often normed below 5 年龄 (例如, 这
Benton Visual Retention Test; Sivan, 1992), or only focus on speciﬁc categories (例如, faces
or shapes; Test of Memory and Learning; Reynolds & Bigler, 1994; Children’s Memory Scale;
科恩, 1997). Another widely used tool focuses on recall of a single complex ﬁgure (IE。,
the Rey-Osterrieth Complex Figure; Osterrieth, 1944; Rey, 1941), and performance is mea-
sured by having children copy the ﬁgure by drawing (例如, Akshoomoff & Stiles, 1995). 这
may underestimate the amount of visual detail that can be remembered, especially in children
younger than 6 年 (Frisk, Jakobson, 骑士, & 罗伯逊, 2005).

Could young children’s visual memories be as ﬁne-grained as those of adults? Or do
children begin with coarser visual memory representations, which then develop in level of
细节? Visual recognition memory in infants has been studied extensively (for review, see Rose,
费尔德曼, & Jankowski, 2004).
In terms of short-term iconic visual memory, 6-month-olds
demonstrate a 5-object memory capacity, which nearly matches adults’ 6-object capacity
(Blaser & Kaldy, 2010). Additional evidence suggests that infants retain at least some degree
of perceptual detail in object representations over longer delays. At 5–6 months of age, 婴儿
are able to recognize, after a delay of two days, which of two abstract patterns they had seen
(费根, 1973). Three-month-olds demonstrate recognition of faces that they have been habitu-
ated to after a delay of 24 小时 (Pascalis, de Haan, 纳尔逊, & de Schonen, 1998). Research us-
ing the deferred imitation paradigm, in which participants reproduce actions using props (例如,
Meltzoff, 1985), has shown that infants as young as 14 months are capable of remembering
在
actions based on a single observation experience (Bauer, Larkina, & Deocampo, 2010).

开放的心态: 认知科学的发现

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
哦
p
米

我
/

我

A
r
t
我
C
e
–
p
d

F
/

2
1
1
4
1
8
6
8
3
0
2
哦
p
米
_
A
_
0
0
0
1
4
p
d

我

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Visual Memory Capacity in Childhood Ferrara et al.

work on episodic memory (Tulving, 1983), 2- to 4-year-olds exhibit impressive abilities to
remember past events, even after a delay of 24 小时 (Simcock & Hayne, 2003). Children as
young as 15 months demonstrate episodic-like memory for past events (Newcombe, Balcomb,
Ferrara, 汉森, & Koski, 2014), and this ability further improves during middle childhood
(Ghetti, Mirandola, Angelini, Corndoli, & Ciaramelli, 2011). Collectively these studies suggest
the possibility that visual memory capacity could be well-developed early in life. 然而,
much of this work has focused on infants and to our knowledge no study has yet attempted
to measure the sheer capacity of visual memory in childhood.

In the present work, we draw upon the experiment of Brady et al. (2008), which leverages
subtle differences in stimuli to provide insights about the capacity of adult visual memory. 这
current work tests 4- and 6-year-olds’ memory for a large set of images (116). We aim to gain
a better understanding of children’s memory capacity by pushing both the number of items
to be remembered (quantity), as well as the speciﬁcity of information remembered per item
(ﬁdelity) (Konkle et al., 2010). As numerous studies have shown that adults have a massive
capacity for visual information, it is possible that this is a fundamental ability present early in
生活. 如果是这样, we predict that visual memory in children will show the signature characteristics of
the mature system (IE。, detailed memory for a large set of items).

METHOD

参加者

Forty-eight children participated in this study. They were divided into two groups: 4-年-
olds (n = 24, mean age = 4;6, age range = 4;0–4;11, 标准差= 0;3, 12 女性) and 6-year-olds
(n = 24, mean age = 6;6, age range = 6;0–6;11, 标准差= 0;4, 12 女性). Two additional
included in the subsequent analyses due to non-
children were tested but were not
参与 (n = 1) or equipment failure (n = 1). The study was approved by the local Insti-
tutional Review Board. Children provided verbal assent for their participation and caregivers
provided written informed consent. All participants had parent-reported normal or corrected-
to-normal vision. Children were recruited via local parent groups and preschools. At the end
of the study, children chose a small toy to take home.

Stimuli and Design

Stimuli were drawn from the image stimulus set of Brady et al. (2008), which includes objects
from distinct basic-level categories in order to minimize conceptual interference (Konkle et al.,
2010; Koutstaal & Schacter, 1997). In some cases particular items were revised in order to
be recognizable to young children (例如, a cupcake instead of a contact lens case). 这些
revised stimuli were gathered via Internet searches using Google Image Search. 完整的
set of stimuli is available on our website (Ferrara, Furlong, 公园, & Landau, 2017A). Sample
stimuli are shown in Figure 1.

Within each age group, eight children were assigned to one of

three conditions
(类别, Exemplar, or State). These conditions were modeled on those of Brady et al. (2008)
and systematically varied the level of detail required for accurate performance at test. 在里面
Category condition, a previously seen item was paired with a new item that was from a dif-
ferent basic-level category (例如, a baby doll vs. a phone). In this condition, remembering the
basic-level category of the previously seen item is sufﬁcient to choose the correct image. 在

开放的心态: 认知科学的发现

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
哦
p
米

我
/

我

A
r
t
我
C
e
–
p
d

F
/

2
1
1
4
1
8
6
8
3
0
2
哦
p
米
_
A
_
0
0
0
1
4
p
d

我

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Visual Memory Capacity in Childhood Ferrara et al.

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
哦
p
米

我
/

我

A
r
t
我
C
e
–
p
d

F
/

2
1
1
4
1
8
6
8
3
0
2
哦
p
米
_
A
_
0
0
0
1
4
p
d

我

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

数字 1. Nine example test pairs presented during the two-alternative forced-choice task for
the three conditions (类别, Exemplar, and State). 有 50 test pairs per condition. 这
complete stimulus set is available at http://parklab.johnshopkins.edu/IMAGE_SETS_2.html.

开放的心态: 认知科学的发现

Visual Memory Capacity in Childhood Ferrara et al.

the Exemplar condition, a previously seen item was paired with a new item from the same
basic-level category (例如, a purple backpack vs. a yellow backpack). Accuracy in this con-
dition would require remembering the particular item that had been seen; remembering only
the basic-level category of the previously seen item would result in chance performance at
测试. 最后, in the State condition, the previously seen item was paired with a new item that
portrayed the same object but in a different pose or state (例如, a whole lemon vs. a sliced
In this condition, memory for just the category or even the speciﬁc identity of the
lemon).
object would be inadequate to make the correct choice.

程序

Images were shown for 3 seconds each.

Children were seated at a laptop computer at a comfortable viewing distance. The study in-
volved two parts: familiarization (一般, 15.35 minutes) and test (一般, 20.10 min-
utes).1 During familiarization, children viewed a continuous stream of 116 images on a
In order to maintain children’s
computer screen.
注意力, a one-back repetition detection task was used (called “catch the sneaky picture”).
Children were instructed to monitor the stream of images for “sneaky” (repeated) 图片
that “took two turns in a row.” They indicated that they saw a “sneaky one” by clapping.
Whenever they identiﬁed a repeated picture they earned a sticker. Children achieved 100%
accuracy in detecting the repeated image. Eight items in the stream were designated as re-
peats, and none of the repeated items were used at test. Repeat images were inserted into
the stream randomly, with the constraint that at least three nonrepeated images must inter-
vene. The remaining 100 nonrepeated images were presented in a randomized order for each
参与者.

Children were not instructed to remember the objects they would see. 相当, 他们
were simply told to “watch for the sneaky picture.” This aspect of the procedure differs
from that of Brady et al.
(2010), in which adult participants were
explicitly informed at the start of the study that they should try to remember all the items
they saw. We opted not to explicitly instruct participants in this way, due to concerns
about variable use of metacognitive strategies by children of this age (Bjorklund, Dukes, &
棕色的, 2009).

(2008) and Konkle et al.

At test, participants completed 50 trials of a two-alternative forced choice decision.
Each item that was viewed during familiarization was paired with a new image, and par-
ticipants were asked to indicate which one they had seen before. After every 10 completed
test trials children earned a sticker. They were permitted to proceed at their own pace. 这
identity of the foil image was determined by the condition assigned to the particular participant
(类别, Exemplar, or State). The test trials were presented in a randomized order for each
参与者.

As in Brady et al.

(2008), 这 50 pairs of images used to test each condition remained
consistent across participants (例如, in the Category condition, the leaf was always paired
with the paint box, the doll was always paired with the phone, ETC。). Within each pair, 这

1 Durations are given as averages because children were permitted to proceed at their own pace. No individual

participant exceeded +/– 4 minutes of the given familiarization and test durations.

开放的心态: 认知科学的发现

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
哦
p
米

我
/

我

A
r
t
我
C
e
–
p
d

F
/

2
1
1
4
1
8
6
8
3
0
2
哦
p
米
_
A
_
0
0
0
1
4
p
d

我

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Visual Memory Capacity in Childhood Ferrara et al.

image assigned to be familiarized and the image assigned to be the test foil were counter-
balanced across participants (例如, some participants were familiarized to the leaf and saw
it paired with the paint box at test, while others were familiarized to the paint box and saw
it paired with the leaf at test). 这 50 familiarized images that were later tested were dis-
tributed evenly throughout the entire stream of images during the familiarization portion of
the study.

结果

Children in both age groups were highly accurate on the forced-choice decision at test
In the Category condition, the mean percent correct
across all three conditions (数字 2).
for 4-year-olds was 95.00% (SE = 0.86%) and the mean percent correct for 6-year-olds was
98.00% (SE = 0.23%). In the Exemplar condition, the mean percent correct for 4-year-olds was
94.00% (SE = 1.00%), and the mean percent correct for 6-year-olds was 91.13% (SE = 1.17%).
In the State condition, the mean percent correct for 4-year-olds was 86.25% (SE = 0.85%) 和
the mean percent correct for 6-year-olds was 87.75% (SE = 0.99%). No gender differences in
accuracy were found for either the 4-year-olds, two-tailed t(22) = –0.35, p = .73, Cohen’s d =
–0.14, or 6-year-olds, t(22) = 0.30, p = .77, Cohen’s d = 0.12.

Participant accuracy (percent correct) was submitted to a 3 (Condition: 类别, Exem-
plar, 状态) X 2 (年龄: 4-year-olds, 6-year-olds) analysis of variance (ANOVA) with Condition and
Age as between-subjects factors. This analysis yielded a main effect of condition, F(2, 42) =
7.05, p = .002, ηp
= .25, indicating that performance was affected by the level of detail

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
哦
p
米

我
/

我

A
r
t
我
C
e
–
p
d

F
/

2
1
1
4
1
8
6
8
3
0
2
哦
p
米
_
A
_
0
0
0
1
4
p
d

我

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

数字 2. Percent correct in the Category, Exemplar, and State conditions for 4-year-olds and
6-year-olds. Error bars reﬂect one standard error of the mean (SE). Chance performance (50%) 是
indicated by the dashed line.

开放的心态: 认知科学的发现

Visual Memory Capacity in Childhood Ferrara et al.

required to remember the familiar item at test. The main effect of Age was nonsigniﬁcant,
F(1, 42) = .08, p = .78, ηp
= .002, as was the interaction between Age and Condition,
2
F(2, 42) = .69, p = .51, ηp
= .03. 因此, performance did not signiﬁcantly improve with age,
and the different conditions did not differentially affect accuracy for 4-year-olds versus 6-year-
olds. Analyses using the Scheffé post hoc criterion for signiﬁcance showed that the Category
condition did not elicit signiﬁcantly higher accuracy than the Exemplar condition, F(2, 42) =
1.57, p = .32, 95% conﬁdence interval, CI = [-.03, .10]. The difference between accuracy
in the Exemplar and State conditions was also not signiﬁcant, F(2, 42) = 2.45, p = .10, 95%
CI = [-.01, .12]. Comparison of the Category and State conditions, 然而, showed that the
Category condition elicited signiﬁcantly higher accuracy than State, F(2, 42) = 6.97, p = .002,
95% CI = [-.03, .16]. The greater accuracy in the Category condition compared to State stands
as a replication of Brady et al.’s (2008) ﬁndings with adults. This indicates that both children and
adults were able to store and retrieve fewer items when speciﬁc states had to be remembered,
compared to simply remembering the object’s category.

One possibility is that the visual similarity of the item pairs in the three conditions
varied such that poorer performance in the State condition was due to higher visual simi-
larity between targets and their foils. We used a phase correlation algorithm to quantify the
relative visual similarity between two test images. We found that the State condition had the
highest overall interitem similarity, but higher image similarity values for speciﬁc item pairs
did not predict behavioral performance in terms of accuracy (see Supplemental Figure 1 和
Supplemental Table 1; Ferrara, Furlong, 公园, & Landau, 2017乙).

We also considered the potential effect of the number of intervening items (“item delay”)
on performance at test. Each participant was exposed to the 100 familiarization images in a
randomized order, 和 50 test pairs were also presented in a randomized order. 从这些
orders we computed the number of intervening items between the point at which an item was
viewed during familiarization and when it was viewed again at test.
If item delay affected
准确性, this would be reﬂected by a decrease in performance with the increasing number
of intervening items. No effects of item delay were found (see Supplemental Figure 2; Ferrara
等人。, 2017乙).

Although the omnibus ANOVA did not reveal a main effect of age, we further explored
potential age effects by conducting linear regression analyses to determine whether exact age
(measured in terms of years, 月, and days) signiﬁcantly predicted accuracy at test. 这些
analyses showed no signiﬁcant effects of age for any of the conditions (see Supplemental
数字 3 and Supplemental Table 2; Ferrara et al., 2017乙). We also considered the number
of test pairs for which children did not obtain 100% accuracy in each condition (see Supple-
mental Figure 4; Ferrara et al., 2017乙). Four- and six-year-olds differed only in the category
状况, where 4-year-olds had signiﬁcantly more test items that fell below 100% 准确性,
χ2

(2, n = 100) = 6.45, p = .01.

Given that both 4- and 6-year-olds are clearly storing a large amount of visual detail
in order to achieve such high levels of performance, we sought to quantify the capacity of
children’s visual memory. As noted by Brady et al. (2008), true estimates of memory capacity
take into account not only the number of items stored, but also the amount of information that
must be remembered per item. To quantify memory capacity, Brady et al.
(2008) 分析过的
their adult data using a model by Landauer (1986). This model is not based on visual similarity,
and instead assigns each picture a random code (b bits long) and calculates the number of bits

开放的心态: 认知科学的发现

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
哦
p
米

我
/

我

A
r
t
我
C
e
–
p
d

F
/

2
1
1
4
1
8
6
8
3
0
2
哦
p
米
_
A
_
0
0
0
1
4
p
d

我

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Visual Memory Capacity in Childhood Ferrara et al.

required to correctly make a decision about which image has been seen and which has not.
Errors occur when two images are mistakenly assigned the same code. The model provides an
estimate of the optimal code length (bits per item) based on the total number of items in the
image set and the percentage correct on a two-alternative forced-choice memory test.2 Adults
in Brady et al. (2008) obtained 93% in the Novel condition (in our experiment called the
Category condition). Using Landauer’s model, Brady et al.
found that the optimal code for
this performance would require 13.8 bits per item. We applied Landauer’s model to our own
data to see if a comparable result would be obtained, despite the fact that the image set of the
present study is much smaller and children could well have a capacity for far fewer bits than
adults. For our Category condition, with an image set of 100 items and 96.5% 正确的 (对全部
children tested), the optimal code for this performance would require 10.43 bits per item. 这
is close to the number of bits per item estimated by Brady et al. for their adult participants.
As Brady et al. 著名的, their own empirically derived value is comparable with estimates of
10–14 bits needed to model adult performance in previous large-scale experiments with adults
(Landauer, 1986). It is even more striking that the estimated bits per item for our child data
falls within this range.

We also modeled the Exemplar and State data following the assumptions laid out by
Brady et al. (2008), which assume a hierarchical organization of memory, where the category
bits appear ﬁrst and additional bits of information then specify the exemplar or state of the
item in that category. The exemplar bits are assumed to be nonoverlapping with the state bits
in the hierarchy. 例如, as Brady et al. (2008) assumed, two additional bits are required
to code a particular exemplar (12.43 bits per item, in comparison to 10.43 bits per item to
code a particular category) and two more are required to code a particular state (14.43 bits
per item). Given that the State condition will require the highest number of bits per item, 我们
can use that number as an estimate of the representational capacity of children’s visual mem-
奥里, assuming that all items are coded by the optimal set of features. That estimate is a massive
number ≈ 22,000 (2
).3 This analysis does not tell us the true visual information capacity of
系统 (Brady et al., 2008), but serves as a formal demonstration of the impressively
large memory capacity that must exist to encode the many unique exemplar and state fea-
tures and obtain the high rates of accuracy observed for both children and adults. 全面的,
these results point to the remarkable ability of 4- and 6-year-olds to represent ﬁne-grained de-
tails of previously seen items, encompassing variation across objects in a single basic-level
类别, and variation within an individual object in that category.

14.43

讨论

The present study demonstrates impressive visual memory performance by children as young
作为 4 年龄, both in terms of the large number of items and the level of detail required
为了认可. Children showed accurate recognition not only for the basic-level category
of the item they saw (类别), but also for unique featural details (Exemplar), and speciﬁc
information about position and arrangement of parts (状态). As was found previously with
adults, performance declined slightly in the State condition (Brady et al., 2008). 记忆

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
哦
p
米

我
/

我

A
r
t
我
C
e
–
p
d

F
/

2
1
1
4
1
8
6
8
3
0
2
哦
p
米
_
A
_
0
0
0
1
4
p
d

我

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

2 The number of bits is related to memory performance by the following equation: b = –log2 [1 – (2p –1)

1/n

where p is the percentage correct and n is the number of items in memory.

3 Two objects (target and foil), 和 14.43 bits/object.

开放的心态: 认知科学的发现

Visual Memory Capacity in Childhood Ferrara et al.

performance between 4- and 6-year-olds was largely similar. Whether the same levels of per-
formance (and hence estimated visual memory capacity) holds for children younger or older
than this age range awaits further study. 全面的, these ﬁndings demonstrate that the ﬁdelity of
children’s visual memory shows the signature characteristics of the mature system—detailed
and highly accurate visual memory that extends over a large set of items.

This work highlights clear developmental precursors to the massive visual memory
capacity that has been shown in studies of adults. 为了确定, those studies included more
比 1,000 images while the present includes only 116. 即使是这样, the abbreviated constraints of
the current study show that children retain details for a large number of
项目, 见过
only for 3 seconds each. The estimated number of bits per item based on Landauer’s model
(Landauer, 1986) was comparable to that which Brady et al. (2008) found for adults. This ﬁnd-
ing is surprising, given the large developmental differences in other aspects of memory (例如,
工作记忆 [Gathercole, 1999; 黑尔, Bronik, & Fry, 1997; Luciana & 纳尔逊, 1998]
and episodic memory [Brainerd, Holliday, & Reyna, 2004; DeMaster & Ghetti, 2013; Lloyd,
Doydum, & Newcombe, 2009]). 尤其, high levels of accuracy were observed even when
children were not instructed to remember anything about the images they saw—during fa-
miliarization they were fully engaged in the “catch the sneaky picture” game. This indicates
that explicit strategies are unlikely to account for performance. These data also suggest that
the retrieval and comparison processes necessary to successfully perform the two-alternative
forced-choice test are relatively well-developed in young children. 此外, it is unlikely
that children were able to linguistically encode nuanced differences among the stimuli within
the brief duration for which each image was displayed. Collectively, these ﬁndings indicate
that the storage of ﬁne-grained visual information is a fairly automatic and rapid process in
early childhood.

In our study, children showed exceptional memory capacity for even the subtle visual
differences presented in the State condition. This observation is consistent with ﬁndings from
infant studies, which indicate that object heterogeneity can affect memory performance. 什么时候
habituated to two objects that differ in color, pattern, and texture, 7-month-olds respond to
a change in the number of objects present. But when habituated to two identical objects,
infants fail to respond to a change in object number (Feigenson & 凯里, 2005). 这表明
that in infancy, the distinctiveness of visual detail is an important element toward forming
separate representations for multiple objects, which holds repercussions for further cognitive
加工, such as representation of numerosity. Additional work has shown that when the
perceptual contrast between objects is highly salient, infants show improved working memory
容量 (Zosh & Feigenson, 2015). The current ﬁndings expand upon this developmental
帐户; here we observe high levels of accuracy, even when interitem similarity is quite
高的. Building upon the attention to visual detail that they demonstrate as infants, 4-年-
olds can form distinct and enduring object representations based on slight variations in visual
appearance.

It is important to note that although object memory is highly detailed and accurate in chil-
dren and adults, this does not hold for other types of visual stimuli, such as images of scenes.
On scene memory tasks, both children and adults make consistent errors that are not faithful
to the original stimulus. Research has shown that children as young as 4 years of age demon-
strate the phenomenon of boundary extension (false memory beyond the edges of a view of
a scene; Kreindel & Intraub, 2016), which has also been shown in adults (Intraub, 2010;

开放的心态: 认知科学的发现

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
哦
p
米

我
/

我

A
r
t
我
C
e
–
p
d

F
/

2
1
1
4
1
8
6
8
3
0
2
哦
p
米
_
A
_
0
0
0
1
4
p
d

我

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Visual Memory Capacity in Childhood Ferrara et al.

Intraub & 理查森, 1989). 因此, the highly detailed nature of children’s visual memory
reported here supports aspects of object recognition, but may not extend to other cognitive
功能, such as scene representation.

These ﬁndings suggest several avenues of further investigation. 第一的, what is the visual
memory capacity of infants? Could it be within the same range shown by the 4- and 6-year-
olds in our study? It may be possible to answer this question by using measures such as looking
time to determine whether items from a relatively large set of objects are later recognized as
different from foils that differ in terms of category, exemplar, and/or state. 第二, do the
effects we’ve observed persist over longer periods of time? To ﬁnd out, one could test chil-
dren’s ability to recognize foils over delays of hours, or even days. Capacity could be further
investigated by increasing the set size above 100 图片. The present data suggest that chil-
dren might obtain high levels of accuracy on even larger sets of images, as analyses of item
delay did not reveal any decrease in accuracy as the number of intervening items increased.

最后, this study raises questions about the neural basis of the ability to form and maintain
visual object representations. This work could be expanded to tasks that include neuroimaging
措施, which could further our understanding of the maturation and functioning of key
regions that underlie human memory, such as the dentate gyrus and hippocampus for pattern
separation and completion (Bakker, Kirwan, 磨坊主, & Stark, 2008; Hunsaker & Kesner, 2013;
Jabès & 纳尔逊, 2015).
总共, our ﬁndings demonstrate that children possess impressive
visual memory capacity even early in life, and provide intriguing avenues for future research.

致谢

We are very grateful to all the families who took part in this study. We thank Matthew Levine
for his technical assistance in developing the experimental MATLAB scripts.

This research was supported by an Integrative Graduate Education and Research Trainee-
ship through the National Science Foundation (DGE 0549379 to K. Ferrara), a T32 Postdoctoral
Research Fellowship through the National Institutes of Health (5T32 HD 046388 to K. Ferrara),
a Johns Hopkins University Provost Undergraduate Research Award (to S. Furlong), a grant from
the National Institutes of Health (NINDS RO1 050876 to B. Landau), and a grant from the
美国国立卫生研究院 (NIH R01 EY026042 to S. 公园).

作者贡献

SP and BL contributed equally to the research design, data analysis, and writing of the paper.
All authors contributed to the study concept and design. Testing and data collection were
performed by KF and SF. KF and SF performed data analysis and interpretation under the super-
vision of BL and SP. KF and SF drafted the manuscript, and BL and SP provided critical revisions.
All authors approved the ﬁnal version of the manuscript for submission.

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
哦
p
米

我
/

我

A
r
t
我
C
e
–
p
d

F
/

2
1
1
4
1
8
6
8
3
0
2
哦
p
米
_
A
_
0
0
0
1
4
p
d

我

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

参考

Akshoomoff, 氮. A。, & Stiles, J.

(1995). Developmental trends in
visuospatial analysis and planning:
我. Copying a complex ﬁg-
乌尔. Neuropsychology, 9, 364–377. 土井:10.1037/0894-4105.
9.3.364

Bakker, A。, Kirwan, C. B., 磨坊主, M。, & Stark, C. 乙.

(2008). Pat-
tern separation in the human hippocampal CA3 and dentate
gyrus. 科学, 319, 1640–1642. 土井:10.1126/science.1152882

Bauer, 磷. J。, Larkina, M。, & Deocampo, J.

(2010). Early memory
发展. 在U. Goswami (埃德。), Blackwell handbook of child-
hood cognitive development (2nd 版。, PP. 253–299). 牛津,
England: 布莱克威尔.

Bjorklund, D. F。, Dukes, C。, & 棕色的, 右. D.

velopment of memory strategies.

(2009). The de-
在米. Courage & 氮. 考恩

开放的心态: 认知科学的发现

Visual Memory Capacity in Childhood Ferrara et al.

(编辑。), The development of memory in infancy and childhood
(PP. 145–175). Hove East Sussex, England: Psychology Press.
Blaser, E., & Kaldy, Z. (2010). Infants get ﬁve stars on iconic memory
测试: A partial-report test of 6-month-old infants’ iconic memory
容量. 心理科学, 21, 1643–1645.

布雷迪, 时间. F。, Konkle, T。, 阿尔瓦雷斯, G. A。, & Oliva, A. (2008). Visual
long-term memory has a massive storage capacity for object
细节. 美国国家科学院院刊, 美国,
105, 14325–14329. 土井:10.1073/普纳斯.0803390105

Brainerd, C. J。, Holliday, 右. E., & Reyna, V. F.

(2004). Behavioral
measurement of remembering phenomenologies: So simple a
child can do it. Child Development, 75, 505–522. 土井:10.1111/
j.1467-8624.2004.00689.x

Castelhano, M。, & Henderson, J. (2005). Incidental visual memory
for objects in scenes. Visual Cognition, 12, 1017–1040. 土井:10.
1080/13506280444000634

科恩, 中号. J.

(1997). Children’s Memory Scale [Administration

手动的]. San Antonio, TX: Psychological Corporation.

DeMaster, D. M。, & Ghetti, S.

(2013). Developmental differences
in hippocampal and cortical contributions to episodic retrieval.
Cortex, 49, 1482–1493. 土井:10.1016/j.cortex.2012.08.004

费根, J. F.

(1973).

Infants’ delayed recognition memory and for-
getting. Journal of Experimental Child Psychology, 16, 425–450.
土井:10.1016/0022-0965(73)90005-2

Feigenson, L。, & 凯里, S. (2005). On the limits of infants’ quantiﬁ-
cation of small object arrays. 认识, 97, 295–313. 土井:10.
1016/j.cognition.2004.09.010

Ferrara, K., Furlong, S。, 公园, S。, & Landau, 乙.

(2017A). Stimulus
sets for “Detailed Visual Memory Capacity Is Present Early in
Childhood.” Park lab. http://parklab.johnshopkins.edu/IMAGE_
SETS_2.html

Ferrara, K., Furlong, S。, 公园, S。, & Landau, 乙. (2017乙). Supplemental
material for “Detailed Visual Memory Capacity Is Present Early
in Childhood.” Open Mind: 认知科学的发现,
2(1), 14–25. 土井:10.1162/opmi_a_00014

Frisk, 五、, Jakobson, L. S。, 骑士, 右. M。, & 罗伯逊, 乙.

(2005).
Copy and recall performance of 6–8-year-old children after
standard vs. step-by-step administration of the Rey-Osterrieth
complex ﬁgure. Child Neuropsychology, 11, 135–152. 土井:10.
1080/092970490911289

Gathercole, S. 乙. (1999). Cognitive approaches to the development
of short-term memory. 认知科学的趋势, 3, 410–419.
Ghetti, S。, Mirandola, C。, Angelini, L。, Corndoli, C。, & Ciaramelli, 乙.
(2011). Development of subjective recollection: 理解
of and introspection on memory states. Child Development, 82,
1954–1969. 土井:10.1111/j.1467-8624.2011.01645.x

黑尔, S。, Bronik, 中号. D ., & Fry, A. F. (1997). Verbal and spatial work-
ing memory in school-age children: Developmental differences
in susceptibility to interference. Developmental Psychology, 33,
364–371.

霍林沃斯, A. (2004). Constructing visual representations of nat-
ural scenes: The roles of short- and long-term visual memory.
实验心理学杂志: Human Perception & Per-
formance, 30, 519–537. 土井:10.1016/j.neubiorev.2012.09.014
(2013). The operation of pattern
separation and pattern completion processes associated with
different attributes or domains of memory. 神经科学 & Bio-
behavioral Reviews, 37, 36–58.

Hunsaker, 中号. R。, & Kesner, 右. 磷.

(2010). Rethinking scene perception: a multisource
Intraub, H.
在乙. Ross (埃德。), The psychology of learning and moti-
模型.
休假: Advances in research and theory (卷. 52, PP. 231–264).
圣地亚哥, CA: Elsevier Academic Press.

Intraub, H。, & 理查森, 中号.

(1989). Wide-angle memories of
实验心理学杂志: 学习,

close-up scenes.
记忆, 和认知, 15(2), 179–187.

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
哦
p
米

我
/

我

A
r
t
我
C
e
–
p
d

F
/

2
1
1
4
1
8
6
8
3
0
2
哦
p
米
_
A
_
0
0
0
1
4
p
d

我

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Jabès, A。, & 纳尔逊, C. A. (2015). 20 years after “The ontogeny of
human memory: A cognitive neuroscience perspective,” where
International Journal of Behavioral Development, 39,
are we?
293–303. 土井:10.1177/0165025415575766

Konkle, T。, 布雷迪, 时间. F。, 阿尔瓦雷斯, G. A。, & Oliva, A. (2010). Concep-
tual distinctiveness supports visual long-term memory for real-
实验心理学杂志: General,
world objects.
139, 558–578. 土井:10.1037/a0019165

Koutstaal, W., & Schacter, D. L. (1997). Gist-based false recognition
of pictures in older and younger adults. Journal of Memory and
语言, 37, 555–583. 土井:10.1006/jmla.1997.2529

Kreindel, E., & Intraub, H.

(2016). Anticipatory scene representa-
tion in preschool children’s recall and recognition memory. 的-
velopmental Science, 20(5), e12444. 土井:10.1111/desc.12444

Landauer, 时间. K.

(1986). How much do people remember? 一些
estimates of the quantity of learned information in long-term
记忆. 认知科学, 10, 477–493.

Lloyd, 中号. E., Doydum, A. 奥。, & Newcombe, 氮. S. (2009). 记忆
binding in early childhood: Evidence for a retrieval deﬁcit. Child
发展, 80, 1321–1328. 土井:10.1111/j.1467-8624.2009.
01353.X

Luciana, M。, & 纳尔逊, C. A. (1998). The functional emergence of
prefrontally-guided working memory systems in four- to eight-
year-old children. Neuropsychologia, 36, 273–293. 土井:10.
1016/S0028-3932(97)00109-7

Meltzoff, A. 氮.

(1985).

Immediate and deferred imitation in
fourteen- and twenty-four-month-old infants. Child Develop-
蒙特, 56, 62–72.

Newcombe, 氮. S。, Balcomb, F。, Ferrara, K., 汉森, M。, & Koski,
J. (2014). Two rooms, two representations? Episodic-like mem-
ory in toddlers and preschoolers. Developmental Science, 17,
743–756. 土井:10.1111/desc.12162

Osterrieth, 磷. A.

(1944). Le test de copie d’une ﬁgure complex:
Contribution a l’e’tude de la perception et de la me’moire [这
complex ﬁgure copy test: Contribution to the study of perception
and memory]. Archives de Psychologie, 30, 286–356.

Pascalis, 奥。, de Haan, M。, 纳尔逊, C. A。, & de Schonen, S. (1998).
Long-term recognition memory for faces assessed by visual
杂志
paired comparison in 3- 和 6- month-old infants.
实验心理学: 学习, 记忆, 和认知, 24,
249–260.

Rey, A. (1941). L’examen psychologique dans les cas d’encepha-
lopathie traumatique [Psychological examination in cases of
traumatic brain injury]. Archives de Psychologie, 28, 286–340.
Reynolds, C. R。, & Bigler, 乙. D. (1994). Test of Memory and Learning

(TOMAL). Austin, TX: PRO-ED.

Rose, S. A。, 费尔德曼, J. F。, Futterweit, L. R。, & Jankowski, J. J. (1997).
Continuity in visual recognition memory: Infancy to 11 年. 在-
telligence, 24, 381–392. 土井:10.1016/S0160-2896(97)90067-2
(2004). Infant visual

Rose, S. A。, 费尔德曼, J. F。, & Jankowski, J. J.

recognition memory. Developmental Review, 24, 74–100.

开放的心态: 认知科学的发现

Visual Memory Capacity in Childhood Ferrara et al.

Simcock, G。, & Hayne, H.

(2003). Age-related changes in verbal
and nonverbal memory during early childhood. Developmental
心理学, 39, 805–814.

Sivan, A. 乙.

(1992). Benton Visual Retention Test (5第三版。). 桑

Antonio, TX: Psychological Corporation.

Standing, L.

(1973). 学习 10,000 pictures. Quarterly Jour-
nal of Experimental Psychology, 25, 207–222. 土井:10.1080/
14640747308400340

Tulving, 乙. (1983). Elements of episodic memory. Oxford psychology