Filetype PDF | Posted on 30 Jan 2023 | 2 years ago
Authentication
digital resources
144x Filetype PDF File size 0.48 MB Source: link.springer.com
Behavior Research Methods
2009, 41 (2), 309-317
doi:10.3758/BRM.41.2.309
Picture span test: Measuring visual
working memory capacity involved in
remembering and comprehension
AZUMI TANABE AND NAOYUKI OSAKA
Kyoto University, Kyoto, Japan
The working memory system is assumed to operate with domain-specific (verbal and visuospatial) resources
that support cognitive activities. However, in research on visuospatial working memory, an appropriate visual
working memory task has not been established. For the present study, a novel task was developed: the picture
span test (PST). This test requires memorizing parts of scene images while comprehending various scene situ-
ations simultaneously. Results of correlation analyses and a factor analysis among college students (n 52)
validated that PST can predict visuospatial cognitive skills whereas a simple visual storage task and a verbal
working memory task cannot. Furthermore, an error analysis indicated that inhibition is important for visuo-
spatial working memory. Additionally, PST is considered to reflect individual differences in the visual working
memory capacity. These findings suggest that the PST is appropriate for measuring visual working memory
capacity and can elucidate its relationship to higher cognition.
Numerous studies of working memory have suggested and two domain-specific storage components—namely,
that the working memory system supports various high- the phonological loop and the visuospatial sketch pad.
level cognitive activities, such as reading comprehension, Consistent with this proposal, Shah and Miyake (1996)
reasoning, language learning, and so on (e.g., Daneman & have demonstrated the separability of spatial and verbal
Carpenter, 1980; Engle, Tuholski, Laughlin, & Conway, working memory resources using the spatial span test.
1999; Kyllonen & Christal, 1990; M. Osaka & N. Osaka, They showed that scores on the spatial span test corre-
1994). The concept of working memory represents a mod- late positively with spatial ability measures but not with
ification and extension of an earlier concept, short-term verbal ability measures. By contrast, RST scores corre-
memory, which assumed a mere passive storage function. late positively with verbal ability measures but not with
Since the concept of working memory is assumed to in- spatial ability measures (N. P. Friedman & Miyake, 2000;
volve executive function, working memory tasks require Handley, Capon, Copp, & Harper, 2002; Miyake, Fried-
processing and maintaining information. Therefore, tasks man, Rettinger, Shah, & Hegarty, 2001; N. Osaka et al.,
measuring the capacity of working memory have higher 2004; Shah & Miyake, 1996). These results lead us to
correlations with cognitive ability measures than those of the conclusion that working memory tasks that require
other traditional short-term memory tasks (for reviews, maintaining materials of a certain domain can predict only
see Daneman & Merikle, 1996). Studies of working mem- cognitive skills of the corresponding domain.
ory frequently use the reading span test (RST), which re- Moreover, some studies have demonstrated a separabil-
quires reading several sentences aloud while maintaining ity of spatial and visual working memory resources by the
a word from each sentence. As a result, the capacity of selective interference with memory performance (Della
working memory measured by span tasks (RST and so on) Sala, Gray, Baddeley, Allamano, & Wilson, 1999; Hecker
is considered to be predictive of cognitive skills. & Mapperson, 1997; Klauer & Zhao, 2004; Tresch, Sin-
However, further problems arise from several findings namon, & Seamon, 1993), by the differentiation of devel-
in research on working memory. In this study, these un- opment (Hamilton, Coates, & Heffernan, 2003; Logie &
settled problems are examined in more detail. Pearson, 1997), and the differentiation of the neural basis
(Courtney, Ungerleider, Keil, & Haxby, 1996; Smith,
The Separability in Working Memory Resources Jonides, Koeppe, Awh, Schumacher, & Minoshima,
The relation between working memory and cognitive 1995); but these studies used only simple storage tasks
abilities has been proposed to involve domain-specific (not working memory tasks—i.e., processing-and-storage
resources (i.e., verbal and visuospatial; Baddeley, 1986). tasks). Thus, it is conceivable that an appropriate working
This is postulated in Baddeleys model of working mem- memory task has not yet been established in the visual
ory, which has a domain-general executive component domain.
A. Tanabe, azumi@l05.mbox.media.kyoto-u.ac.jp
309 © 2009 The Psychonomic Society, Inc.
310 TANABE AND OSAKA
Considering the domain-specific character of working ing memory task and a visual simple storage task were
memory resources, it is necessary to examine features of included in this study. We used psychometric subtests of
the working memory system in each domain separately, visuospatial abilities from Kyoto University NX15 intel-
especially the visual domain. ligence test (R. Osaka & Umemoto, 1984), a standard-
ized intelligence test widely used in Japan, as visuospa-
The Differentiation Between Working Memory tial cognitive ability measures, and, as a verbal working
Tasks and Short-Term Memory Tasks memory task, we used the Japanese reading span test
Studies of verbal working memory have suggested that (M. Osaka & N. Osaka, 1994). The visual simple stor-
working memory tasks such as RST better predict cogni- age task required memorizing images of familiar objects.
tive ability measures than do traditional short-term mem- This task was called the simple object span test (SOST)
ory tasks, such as the word span test (memorizing word and was designed as a visual analogue of the word span
lists), the digit span test (memorizing a set of numbers), test, which typically has lower correlations with language
and so on (Daneman & Merikle, 1996). In other words, the comprehension measures (Daneman & Carpenter, 1980).
predictability of cognitive abilities mainly characterized We predicted that the score of PST, our novel task, should
working memory tasks to indicate that the tasks are valid show higher correlations with the performances of visuo-
measures of working memory, not of short-term memory. spatial cognitive tasks than the verbal working memory
However, few studies on visuospatial working memory task (RST) and the visual simple storage task (SOST)
have not shown differences in cognitive involvement be- would, if PST truly reflects the capacity of visual working
tween a working memory task and a simple storage task; memory closely related to visuospatial cognition.
that is, correlations between the spatial span test and spa-
tial ability measures did not differ from those of a simple The Relationship With Individual Differences
spatial storage task (Shah & Miyake, 1996), whereas To examine the skills involved in the relation between
verbal working memory studies to date showed that the the capacity of visual working memory and cognitive ac-
working memory task, RST, has higher correlations with tivities, the errors in PST were investigated. In the present
cognitive tasks than the simple word storage task does, study, we focused on intrusion errors and token errors.
indicating that a parallel differentiation of the crucial fea- Intrusion errors are defined as false recall or recognition
ture of working memory has not been ascertained for the of information presented simultaneously with target infor-
visuospatial domain. This differentiation must be obtained mation. Several studies of verbal working memory have
with an appropriate task to measure the capacity of visual analyzed intrusion errors (e.g., De Beni, Palladino, Paz-
working memory. zaglia, & Cornoldi, 1998; M. Osaka, Nishizaki, Komori,
In light of the problems described above, the primary & N. Osaka, 2002). The number of intrusion errors has
goal of the present research was to develop a new visual been considered an index of poor inhibition (the ability
working memory task offering greater prediction of re- to suppress irrelevant information) in comprehension (De
lated cognitive activities than a simple storage task. Beni et al., 1998; Gernsbacher, Varner, & Faust, 1990). In-
hibition is assumed to be included in executive functions
A Novel Visual Working Memory Task: of working memory (Miyake et al., 2000), so we predicted
The Picture Span Test that intrusion errors in PST would significantly correlate
To solve the problems discussed above, we developed with performance. By contrast, token errors were defined
a new visuospatial working memory task named the pic- as false recognition of a novel object that belongs to the
ture span test (PST) in order to measure the capacity to same category as the target object (Hollingworth, 2006).
“read” and maintain visual information. Corresponding If a participant memorized target objects as words (i.e.,
to reading sentences in RST, PST entails seeing visual objects names), rather than as visual information, the
contexts—that is, understanding the situation in the scene. number of token errors would increase, so token errors
Accordingly, PST uses scene images containing familiar were not expected to reflect performance. In view of these
objects within context-rich surroundings. In addition to predictions, a secondary goal of the present research was
remembering a designated target object, participants are to demonstrate that inhibition plays a crucial role in visual
required to judge whether each object in the scene is likely working memory, but verbal coding does not.
to appear in the situation of the background (congruent) Finally, our advanced goal concerns individual differ-
or not (incongruent). This procedure was modeled after ences in visual working memory. It has been found that
a variant of RST, the silly sentence span test, which re- individual differences in working memory capacities can
quires judging each sentence to be semantically true or account for many aspects of individual differences in
false in order to confirm that participants actually read cognitive skills (Baddeley et al., 1985; Just & Carpenter,
it (Baddeley, Logie, & Nimmo-Smith, 1985). Since such 1992). Likewise, it is readily supposed that the capacity
scene perception has been considered “comprehension” of visual working memory should reflect individual dif-
(A. Friedman, 1979), we reasoned that PST would predict ferences. We aimed to ascertain whether PST allows a sig-
relevant cognitive abilities as well as RST does. nificant distinction between individuals with high visual
To confirm the validity of the novel task, we investi- working memory spans and those with low spans.
gated whether the score on PST positively correlated with These analyses may demonstrate the relationship be-
visuospatial ability measures. In addition, a verbal work- tween visual working memory capacity and executive
PICTURE SPAN TEST FOR VISUAL WORKING MEMORY 311
function and indicate that PST has the potential to be a example of Figure 1A, the leftmost part of the third image repre-
valuable tool for future research on individual differences sents the part of the token object, a novel chair), another part of the
of visual working memory. original scene image (in the example of Figure 1A, the second part
on the left of the third image and the third part on the left of the
fourth image represent such parts), a part of a novel object, and the
METHOD asterisk to select in case there was not a correct answer among the
choices (the third image of Figure 1A applies in this case). Images
Participants of choices were randomly located, except for the asterisk, which
Fifty-two graduate and undergraduate students at Kyoto Uni- was located at the far right of the choices (see the third and fourth
versity (35 male, 17 female; mean age, 24.3; SD, 3.35; age range, images of Figure 1A).
20–37 years) participated as volunteers. All participants reported SOST. This task was designed to measure the visual storage ca-
normal or corrected-to-normal vision and no color vision defects. pacity of traditional short-term memory. Participants were presented
with a set of images of a familiar object with a uniform background
Materials, Design, and Procedure of olive green. None of the target objects contained this color, to
Participants performed three memory span tasks (PST, RST, and avoid their blending into the background. This method followed
SOST) and three psychometric subtests from Kyoto Universitys Hollingworth (2006). Participants were required to count aloud re-
NX15 intelligence test (R. Osaka & Umemoto, 1984): the plate peatedly from 1 to 9 as a means of the articulatory suppression.
matching test, the paper unfolding test, and the figure reconstruc- Similarly to PST, four choices for serial recognition were presented
tion test. Psychometric subtests were administered using paper and for 1 sec after each set: the target object (or the token object), two
pencils. PST and SOST were performed on a Windows XP per- novel objects, and the asterisk. Other procedures were the same as
sonal computer using Visual Basic 6.0. The images for PST and in PST (for an example of SOST, see Figure 1B).
SOST were newly generated by Shade 7.5 (commercial 3-D graphic RST. This task was designed to measure the capacity of verbal
software) and connected content libraries (e-frontier, Inc., Tokyo) working memory. Participants were required to read aloud a set of
as rendered 3-D computer graphic images. Each image for SOST unrelated Japanese sentences while concurrently remembering an
contained a daily object (see Figure 1B for examples), and each underlined word (the target word) in each sentence. After reading the
image consisted of 3–4 objects and a scenery background (e.g., sentences in each set, participants were requested to orally recall, in
room, coast, street, park, and so on; see Figure 1A for examples the correct serial order, all of the target words of the given set. RST
of scene images). The scene images subtended a visual angle of was administered using standard procedures (Daneman & Carpen-
24.6º 18.6º (640 480 pixels), and the object images subtended ter, 1980; M. Osaka & N. Osaka, 1994).
an angle of 6.1º 6.1º (150 150 pixels) at the 45-cm viewing Next, three visuospatial subtests of the Kyoto University NX15
distance. Seventy images of scenes in PST and 70 images of objects intelligence test were used to assess visuospatial cognitive abilities.
in SOST were used. Each image was used once in all trials. RST The plate matching test. This task required imagining a digit
was done using a booklet of index cards with printed sentences. The plate and the character plate and rotating the plates mentally. Each
set size of each memory span test was increased from 2 to 5, and plate consisted of a matrix (four rows and four lines). In each cell
in each set size five trials were performed, as usual. The order of of the digit plate there was a digit from 1 to 16, and in each cell of
memory span tests and psychometric tests was pseudorandomized the character plate there was a hiragana (Japanese cursive syllabary)
and counterbalanced across participants. Detailed procedures of character, chosen from among i, ro, ha, ni, ho, he, to, chi, ri, nu, ru,
each task are described below. wo, wa, ka, yo, and ta (the order of these characters is equivalent to
PST. This task was designed to assess visual working memory the alphabetic order of old Japanese). When the mobile digit plate is
capacity. Participants were presented with a set of scene images for put on the immobile character plate, matching between the character
4 sec. The task required judging whether each object was congruent and the digit is performed mentally to obtain answers (3 min for
with the semantic constraints imposed by the background of each 12 subitems; see Figure 2A). In this example, participants need to
scene image, while concurrently remembering the part of the target say which character was under the digit 4).
surrounded with a red square frame; see the top and second images The paper unfolding test. This task required imagining a square
of Figure 1A, in which the red frames are represented as gray square sheet of paper folded in two, three, or four, and imagining making
frames in the scene images. The top scene contains congruent ob- several holes in the sheet. When the sheet was unfolded, the loca-
jects: a desk, a chair, and a lamp, with the background “a room,” tion of holes in the sheet of paper was to be identified (1 min for
and the second scene contains an incongruent object: a traffic sign. 12 subitems; see Figure 2B). In this example, participants had to
The red frame that indicated the target part subtended a visual angle imagine that a square sheet of paper folded in two had had two holes
of 6.1º 6.1º. The frame was presented for the final 1 sec of the made in it, then they had to choose the correct illustration of the
duration in order to prevent participants from memorizing the target sheet unfolded.
without judging the context. Participants were instructed to respond The figure reconstruction test. This task required imagining
by clicking the button placed on the computer screen. Half of all a figure divided and combining the fragments to reconstruct the
scene images contained incongruent objects (one incongruent ob- square. The line to divide the figure was to be drawn (2 min for
ject per image). The evaluation of congruence between an object 13 subitems; see Figure 2C). In this example, the triangle had to be
and a background was conducted via questionnaire in a pilot study. divided and combined into a square. Participants were instructed to
The order of images was randomly assigned. As the articulatory draw the correct dividing line to reconstruct the square.
suppression, participants were asked to count aloud repeatedly from
1 to 9 during the presentation of scene images, in order to inter- Data Analyses
fere with verbal encoding—that is, the strategy by which the target Scoring and evaluating span measures. The score of each span
was remembered as the object name without visual features. As a test was defined as the highest set size for which a participant could
method of the articulatory suppression, the use of counting was correctly answer three of five trials; but if a participant correctly re-
based on the study by Baddeley and Andrade (2000). After the last membered only two of five trials, the participant was given a score of
image of each set was presented, four choices from each image for 0.5 for that set size. For example, if a participant was correct on three
serial recognition were presented for 1 sec, and participants were of the five four-sentence trials, the participant was assigned a span
required to click on the correct answers (see the third and fourth of 4. If the participant was correct on only two of the five trials, the
images of Figure 1A; these images are choices for recognition). participant was assigned a span of 3.5. This method of scoring has
The choices were composed of the target part or the token (a main been used in several previous studies (e.g., M. Osaka & N. Osaka,
object in the target part was replaced by the token object; in the 1994; Shah & Miyake, 1996).
312 TANABE AND OSAKA
A Picture Span Test
4 sec
4 sec
1 sec
1 sec
B Simple Object Span Test
1 sec
1 sec
1 sec
1 sec
Figure 1. Examples of PST and SOST. (A) PST at Set Size 2. In the
present experiment, stimuli were presented in color. (B) SOST at Set
Size 2. Stimuli of this test in the present experiment were also presented
in color.
no reviews yet
Please Login to review.
