Automatic semantic encoding in verbal short-term

advertisement
THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014
http://dx.doi.org/10.1080/17470218.2014.966248
Downloaded by [the Bodleian Libraries of the University of Oxford] at 08:54 11 November 2014
Automatic semantic encoding in verbal short-term
memory: Evidence from the concreteness effect
Guillermo Campoy1, Judit Castellà2, Violeta Provencio1, Graham J. Hitch3, and
Alan D. Baddeley3
1
Faculty of Psychology, University of Murcia, Murcia, Spain
Faculty of Psychology, Universitat Autònoma de Barcelona, Barcelona, Spain
3
Department of Psychology, University of York, York, UK
2
The concreteness effect in verbal short-term memory (STM) tasks is assumed to be a consequence of
semantic encoding in STM, with immediate recall of concrete words benefiting from richer semantic
representations. We used the concreteness effect to test the hypothesis that semantic encoding in standard verbal STM tasks is a consequence of controlled, attention-demanding mechanisms of strategic
semantic retrieval and encoding. Experiment 1 analysed the effect of presentation rate, with slow presentations being assumed to benefit strategic, time-dependent semantic encoding. Experiments 2 and 3
provided a more direct test of the strategic hypothesis by introducing three different concurrent attention-demanding tasks. Although Experiment 1 showed a larger concreteness effect with slow presentations, the following two experiments yielded strong evidence against the strategic hypothesis. Limiting
available attention resources by concurrent tasks reduced global memory performance, but the concreteness effect was equivalent to that found in control conditions. We conclude that semantic effects in
STM result from automatic semantic encoding and provide tentative explanations for the interaction
between the concreteness effect and the presentation rate.
Keywords: Verbal short-term memory; Concreteness effect; Immediate serial recall; Presentation rate;
Dual task paradigm.
Studies of verbal short-term memory (STM)
usually involve the immediate serial recall of lists
of unrelated words or letters. Memory performance
in these situations has been found to be highly
affected by the phonological properties of the tobe-remembered stimuli, as occurs in the similarity
effect (Conrad & Hull, 1964) and in the word
length effect (Baddeley, Thomson, & Buchanan,
1975). As a consequence, it has been traditionally
assumed that verbal information is phonologically
encoded in STM, with participants’ recall relying
on some kind of phonological representations of
the stimuli (Baddeley & Hitch, 1974).
The observation of a number of nonphonological effects in verbal STM tasks, however, suggests
the participation of factors beyond the mere maintenance of phonological traces. Some of these
operate at the level of the individual item. These
include the lexicality effect (better immediate recall
of words than nonwords; Hulme, Maughan, &
Brown, 1991) and the frequency effect (better
immediate recall of high-frequency words than
Correspondence should be addressed to Guillermo Campoy, Universidad de Murcia, Facultad de Psicología, Campus de
Espinardo, 30100, Murcia, Spain. E-mail: gcampoy@um.es
This study was supported by the Spanish Ministry of Science and Innovation [Projects PSI2009-07374 and CSD200800048].
© 2014 The Experimental Psychology Society
1
Downloaded by [the Bodleian Libraries of the University of Oxford] at 08:54 11 November 2014
CAMPOY ET AL.
low-frequency words; Hulme et al., 1997). There is
also abundant evidence that verbal STM may be
influenced by semantic factors (Campoy &
Baddeley, 2008; Haarmann & Usher, 2001;
Martin, 2005; Poirier & Saint-Aubin, 1995;
Walker & Hulme, 1999). A crucial finding supporting the participation of semantic codes in
STM is the concreteness effect, the observation that
immediate serial recall is better for concrete,
high-imageability words such as pencil than for
abstract, low-imageability words like method
(Walker & Hulme, 1999). This effect is assumed
to be a direct consequence of the participation of
semantic codes in STM tasks, with concrete
words benefiting from richer and more distinctive
semantic representations (Acheson, Postle, &
MacDonald, 2010; Allen & Hulme, 2006;
Romani, McAlpine, Martin, 2008; Walker &
Hulme, 1999).
Although the participation of semantic traces in
verbal STM tasks has only been broadly recognized
recently, early evidence was already reported in the
sixties and early seventies by Baddeley and coworkers (Baddeley, 1966; Baddeley & Ecob,
1970; Baddeley & Levy, 1971). However, this evidence was later overshadowed by extensive data
supporting the main role of phonological coding.
Baddeley (1966) examined the immediate serial
recall of lists of five similar and dissimilar words,
with similarity being either phonological (e.g.,
man, mat, map) or semantic (e.g., huge, big, wide).
He found a large effect of phonological similarity
(the standard similarity effect), but also a small
though significant effect of semantic similarity,
suggesting the participation of semantic codes.
Likewise, a subsequent study by Baddeley and
Levy (1971) found effects of semantic similarity
in immediate serial recall of noun–adjective pairs,
provided the pairs were semantically compatible
(e.g., priest–devout). In another study using semantically compatible or incompatible triplets (e.g., my
fine wine vs. wine my fine), Baddeley and Ecob
(1970) found effects of both phonological and
semantic similarity, with the effects of semantic
similarity predominating after a delay. In a recent
review of this early evidence, Baddeley (2012) concludes that performance in verbal STM tasks may
2
rely on both phonological and semantic encoding.
Phonological encoding is rapid and attentionally
undemanding, but phonological traces are fragile
and readily forgotten. In contrast, semantic encoding in standard tasks (i.e., immediate serial recall of
unrelated words) is harder and takes longer to set
up, but traces are more durable.
The idea that semantic encoding in standard
STM tasks takes longer to be set up is congruent
with an early proposal by Shulman (1970, 1971,
1972). According to Shulman, both phonological
and semantic encoding are possible in verbal
STM tasks, but they have different temporal
courses. In standard STM tasks, participants tend
to encode information phonologically rather than
semantically because phonological encoding is
faster and thus more appropriate under the temporal pressure of relatively high presentation rates.
A direct prediction of Shulman’s hypothesis is
that semantic encoding in STM would benefit
from slower presentation rates. To test this prediction, Shulman (1970) analysed the effect of presentation rate in a probe-recognition task. Each trial
involved the presentation of a list of words (at a
presentation rate of a word every 350, 700, or
1400 ms) followed by an instructing cue and a recognition probe. The instructing cue indicated the
kind of response that was required in each particular trial. When the cue was the letter H, participants had to indicate whether the probe was a
homonym of any word in the list (homonymprobe condition); when the cue was the letter M,
participants had to indicate whether the probe
had the same meaning as any presented word
(semantic-probe condition); finally, when the cue
was the letter I, participants had to indicate
whether the probe was identical to any of the list
words (identical-probe condition). Results showed
that recognition of semantic probes improved
with slower presentation rates, whereas a contrary
tendency was found in the homonym and identical
conditions. Shulman interpreted this as supporting
his claim that semantic encoding in STM is time
dependent and, thus, benefits from slower presentation. The procedure in this study, however,
makes strong conclusions difficult. On the one
hand, it seems probable that the task itself
THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014
Downloaded by [the Bodleian Libraries of the University of Oxford] at 08:54 11 November 2014
AUTOMATIC SEMANTIC ENCODING
induced semantic encoding, since participants
knew in advance that they could be prompted to
respond on the basis of meaning. This fact could
lead to some forms of controlled semantic processing that are not present in standard situations.
On the other hand, word lists in Shulman’s experiments involved 10 words, which exceeds the
capacity of STM (Cowan, 2005). This fact raises
the question of to what extent his results are representative of STM performance.
In light of the recent interest in semantic encoding in STM, the idea that presentation rate in STM
tasks determines semantic encoding deserves
renewed attention. If semantic encoding in STM
tasks benefits from a slower presentation rate, this
could suggest the participation of mechanisms
that go beyond the mere automatic activation of
semantic information. Research on the retrieval of
word meaning supports the distinction between a
fast, automatic activation of semantic representations and a slower, more controlled, and effortful
mechanism of strategic retrieval (Badre & Wagner,
2002; Gold et al., 2006; Whitney, Grossman, &
Kircher, 2009). There is a range of situations in
which automatic activation is not sufficient, and a
strategic retrieval is required. It has been suggested
that one of these situations is when the meaning of
a word has to be retrieved in the absence of the contextual support provided by preceding semantically
related words (Whitney et al., 2009). This proposal
is congruent with the idea that semantic encoding
in STM tasks could rely on strategic time-dependent mechanisms, since the absence of contextual
semantic support characterizes standard STM
tasks involving the presentation of lists of unrelated
words. Besides, studies with neuropsychological
patients showing significant reduction of semantic
effects in STM tasks reveal that these patients
have lesions in the left inferior prefrontal cortex
(LIPFC; Martin, 2005), a brain region that has
been associated with top-down control of semantic
memory, including controlled semantic retrieval
and postretrieval selection (Badre, Poldrack, PareBlagoev, Insler, & Wagner, 2005; Badre &
Wagner, 2002; Wagner, Paré-Blagoev, Clark, &
Poldrack, 2001; Whitney, Kirk, O’Sullivan,
Lambon Ralph, & Jefferies, 2011). Therefore,
semantic STM deficits in this kind of patient
could be a consequence of the disruption in mechanisms of control of semantic retrieval (Hoffman,
Jefferies, & Lambon Ralph, 2011), consistent
with the view that semantic encoding in STM
tasks relies on controlled processes.
A further step involving the participation of
controlled mechanisms in verbal STM tasks is the
possibility that participants engage in elaborative
strategies of semantic encoding in order to
improve their performance. Such semantic strategies could involve, for example, the establishment
of semantic links between words and the generation
of stories or visual scenes. Evidence for the importance of these strategies of elaborative encoding
emerges from studies showing that the adoption
of this kind of semantic strategy can eliminate phonological effects usually considered the hallmark of
verbal STM, such as the word-length effect or the
phonological similarity effect (Campoy &
Baddeley, 2008; Logie, Della Sala, Laiacona,
Chalmers, & Wynn, 1996).
The present study aimed to test the hypothesis
that semantic encoding in standard STM tasks
relies on controlled and effortful mechanisms of
strategic semantic retrieval and encoding. To test
this strategic hypothesis, Experiment 1 analysed
how a semantic STM effect, the concreteness
effect, was affected by the stimulus presentation
rate. If semantic encoding depends on effortful,
time-dependent mechanisms, the prediction
would be greater concreteness effect with slower
presentations. Subsequent experiments provided a
more direct test of the strategic hypothesis by
including secondary attention-demanding tasks
together with the STM task. According to the strategic hypothesis, the introduction of secondary
tasks would limit the attentional resources available
for controlled semantic mechanisms and, thus,
would eliminate or at least reduce the concreteness
effect.
EXPERIMENT 1
Experiment 1 aimed to evaluate the relationship
between presentation rate and semantic encoding
THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014
3
CAMPOY ET AL.
Downloaded by [the Bodleian Libraries of the University of Oxford] at 08:54 11 November 2014
in a standard STM task by analysing the effect of
presentation rate on the concreteness effect. On
the basis of the strategic hypothesis, the prediction
was that concrete words would be better recalled
with a slower presentation rate, resulting in larger
concreteness effect at this rate.
Method
Participants
Forty-eight undergraduate students from the
University of Murcia (Spain) took part in the experiment for course credit. All participants in this and
the following experiments were native Spanish
speakers.
Stimuli and apparatus
Two sets of 13 Spanish words, one comprising concrete, high-imageability words, the other comprising abstract, low-imageability words, were chosen
on the basis of the concreteness and imageability
values provided by the LEXESP database
(Sebastián, Martí, Carreiras, & Cuetos, 2000). All
of the stimuli were trisyllabic nouns with the stress
on the penultimate syllable. The sets were matched
for word frequency, familiarity, and number of phonemes (see the Appendix). Concrete words were
bigote, guitarra, molino, cerveza, corbata, espalda,
factura, orquesta, palacio, retrato, alfombra, incendio,
and serpiente (moustache, guitar, windmill, beer,
necktie, back, invoice, orchestra, palace, portrait,
carpet, fire, and snake). Abstract words were
jugada, pureza, suceso, tamaño, dominio, esencia,
indicio, letargo, mandato, ventaja, ambiente, contexto,
and nostalgia (play/move, purity, event, size, dominion, essence, indication, lethargy, mandate, advantage, ambience, context, and nostalgia).
All the words, spoken in a neutral tone by a
female speaker, were digitally recorded and segmented into individual sound files using audio
recorder and editor software. This software was
also employed to adjust the duration of the audio
files (without altering the pitch), so that all the
stimuli lasted 696 ms, which was the mean duration of the original recorded words.
A computer programme generated by E-Prime
2 (Schneider, Eschman, & Zuccolotto, 2002)
4
controlled the experiment. A headset with microphone was used to present stimuli and record
verbal responses for later processing.
Procedure
Participants in this and the following experiments
were tested individually in soundproof booths.
The experiment comprised 52 trials divided into
four blocks, one block for each combination of
word type (concrete, abstract) and presentation
rate (one item per second, one item every two
seconds). Two participants were randomly assigned
to each of the 24 (4!) possible orders of presentation
of these four blocks. The order of presentation of
concrete and abstract lists within a block was determined at random.
Each trial began with the presentation of a row
of dashes, which remained on screen until the participant initiated the trial by pressing the computer
mouse button. Two seconds later, a plus sign
appeared on the computer screen for one second
followed by a five-word list. Words were presented
through the headset at a rate of either one item per
second or one item every two seconds. After the last
stimulus, a question mark was presented on the
screen prompting participants to recall the words
in serial order. They were instructed to substitute
the word espacio (blank) for any word they could
not recall. There was a 16-s time limit to complete
spoken recall.
The 13 lists of each block were constructed for
each participant by selecting words from the pertinent set at random, with the constraint that every
word appeared in five trials, once in each of the
five possible positions within the lists.
Experimental trials were preceded by six practice
trials, three trials at each presentation rate. These
practice trials were constructed so that all the experimental stimuli appeared at least once. A brief rest
period was established before each experimental
block.
Results
Participants’ responses were categorized into the
following three categories: correct responses
(correct words in the right position), order errors
THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014
AUTOMATIC SEMANTIC ENCODING
Table 1. Percentage of correct responses, order errors, and item errors in Experiment 1
Word type
Presentation rate
Concrete
1 s/item
2 s/item
1 s/item
2 s/item
Abstract
Correct responses
Order errors
Item errors
75.29 (16.30)
79.62 (12.56)
72.24 (16.31)
72.72 (13.54)
14.40 (10.77)
12.29 (8.19)
14.66 (10.24)
13.73 (8.36)
12.95 (10.47)
9.65 (7.84)
16.35 (11.81)
16.15 (10.26)
Downloaded by [the Bodleian Libraries of the University of Oxford] at 08:54 11 November 2014
Note: Standard deviations in parentheses.
Table 2. Statistical results for the 2 (word type) × 2 (presentation rate) within-subjects ANOVA on the percentage of correct responses, order
errors, and item errors in Experiment 1
F
df
MSE
p
h2p
17.419
2.408
4.725
1, 47
1, 47
1, 47
68.007
115.211
37.566
,.001
.127
.035
.270
.049
.091
1.128
1.391
0.725
1, 47
1, 47
1, 47
30.619
80.433
23.173
.294
.244
.399
.023
.029
.015
30.608
3.857
6.839
1, 47
1, 47
1, 47
38.451
37.954
16.957
,.001
.055
.012
.394
.076
.127
Effect
Correct responses
Word type
Presentation rate
Word Type × Rate
Order errors
Word type
Presentation rate
Word Type × Rate
Item errors
Word type
Presentation rate
Word Type × Rate
Note: ANOVA = analysis of variance.
(words presented in the current list but recalled in
the wrong position), and item errors, with this
last category subsuming omissions (blank
responses), intraexperimental intrusions (words in
the experimental sets but not presented in the
current list), and extraexperimental intrusions
(words that were not in the experimental sets).
Percentages of responses within each category are
shown in Table 1. To control for different conditions differing in the number of items recalled,
percentages of order errors were calculated with
respect to the number of list items recalled, rather
than to the total number of responses (Murdock,
1976). Table 2 shows the statistical results of the
analyses described below.
Percentages of correct responses were submitted
to a 2 × 2 within-subjects analysis of variance
(ANOVA) with word type (concrete, abstract)
and presentation rate (one item per second, one
item every two seconds) as factors. There was a
main effect of word type, revealing that concrete
words were better recalled than abstract words
(concreteness effect = 4.97%). The main effect of
presentation rate was not significant, but there
was an interaction between word type and presentation rate. This interaction can be attributed to a
larger concreteness effect with the slow presentation rate (concreteness effect = 6.89%) than
with the fast rate (concreteness effect = 3.04%).
The analysis of simple effects revealed that the concreteness effect was significant with slow presentations, F(1, 47) = 22.360, MSE = 50.966, p , .001,
h2p = .322, and fast presentations, F(1, 47) =
4.074, MSE = 54.608, p = .049, h2p = .080. On
the other hand, the presentation rate had no
effect on abstract words (F , 1), whereas concrete
words were better recalled with slow presentations
than with fast presentations, F(1, 47) = 6.101,
THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014
5
Downloaded by [the Bodleian Libraries of the University of Oxford] at 08:54 11 November 2014
CAMPOY ET AL.
MSE = 73,657, p = .017, h2p = .115. The larger
concreteness effect at the slow rate, therefore, can
be specifically attributed to the effect of presentation rate on the recall of concrete words.1
An equivalent ANOVA on the percentages of
order errors showed no significant main effects or
interaction. In contrast, an ANOVA on the percentages of item errors showed a main effect of word
type (more item errors with abstract words) and a
significant interaction between word type and presentation rate (greater effect of word type with slow
presentations). Resembling the pattern found for
correct responses, the analysis of simple effects
revealed that the effect of word type was significant
with both slow presentations, F(1, 47) = 39.377,
MSE = 25.799, p , .001, h2p = .456, and fast presentations, F(1, 47) = 9.355, MSE = 29.609,
p = .004, h2p = .116. In turn, the effect of presentation rate was significant for concrete words, F(1,
42) = 11.543,
MSE = 22.653,
p = .001,
h2p = .197, but there was no effect for abstract
words (F , 1).
Discussion
Results showed a straightforward pattern. Concrete
words were better recalled at the slow rate (a word
every two seconds) than at the fast presentation rate
(one word per second), whereas the recall of
abstract words was not affected by the presentation
rate. This pattern resulted in a larger concreteness
effect with slow presentations, congruent with
Shulman’s proposal that semantic encoding in
STM benefits from a slow presentation rate
(Shulman, 1970, 1971, 1972).
Memory for item identity and serial order in
verbal STM tasks is generally assumed to rely on
mechanisms that are to some extent independent
(Burgess & Hitch, 1999; Murdock, 1976; Nairne
& Kelley, 2004). In the present experiment, the
concreteness effect was specifically a consequence
of more item errors with abstract words than with
concrete words, suggesting that word concreteness
had an effect on memory for item identity rather
than for order information. This fact is congruent
with the widespread opinion that lexical–semantic
factors affect item memory rather than order
memory in STM tasks (Majerus, 2009).
Results of Experiment 1 have diverse implications. On the one hand, they show the importance of taking presentation rate into
consideration whenever semantic encoding in
STM can play a role, with the use of different
rates limiting the comparison between studies.
On the other hand, and more importantly, results
are congruent with the strategic hypothesis according to which semantic encoding in standard STM
tasks relies on controlled, time-dependent mechanisms of strategic semantic retrieval and encoding. It
is important to note, however, that the effect of
presentation rate constitutes only indirect evidence
for the strategic hypothesis, with alternative
interpretations being possible (see General
Discussion). To provide a more direct test of the
strategic hypothesis, the following two experiments
investigated the effect of introducing secondary
attention-demanding tasks on the immediate
serial recall of concrete and abstract words.
EXPERIMENT 2
In Experiment 2, the availability of attentional
resources for strategic semantic processing was
limited by introducing an attention-demanding
random-generation task (Towse & Neil, 1998).
1
In this and the following two experiments, we performed additional ANOVAs on the percentages of correct responses after
including serial position as a within-subject factor. The standard serial position effect, consisting of better recall of the initial
(primacy effect) and last (recency effect) items, was found in the three experiments. In Experiment 1, there was also a significant interaction between presentation rate and serial position, revealing that presentation rate had a greater effect on the recall of items presented
in central positions. The three-way interaction between word type, presentation rate, and serial position was not significant. In
Experiments 2 and 3, there was a significant interaction between concurrent task and serial position, which was a consequence of
greater effect of the concurrent tasks for items presented in central serial positions. Importantly, however, the interaction between
word type, concurrent task, and serial position was not significant in either experiment. For the sake of brevity, and given that
these results do not contribute additional relevant information, they are not further discussed.
6
THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014
Downloaded by [the Bodleian Libraries of the University of Oxford] at 08:54 11 November 2014
AUTOMATIC SEMANTIC ENCODING
More specifically, we introduced a random time
interval generation task (Vandierendonck, De
Vooght, & Van der Goten, 1998), in which participants had to press a button at variable intervals,
trying to generate the most irregular and unpredictable patterns they could produce. In this randomtapping condition, participants presumably had to
allocate attention to monitor the sequence they
were generating, inhibiting repetitions and trying
to maximize variability (Vandierendonck et al.,
1998). Results in this condition were compared to
those in a simple-tapping control condition in
which participants were asked to click the mouse
button at a steady, uniform pace throughout the
presentation of to-be-remembered items. On the
basis of the strategic hypothesis, random tapping
was expected to eliminate or reduce the concreteness effect by limiting the attention resources available for controlled semantic processes.
Method
Participants
Forty-eight undergraduate students from the
University of Murcia participated for course credit.
Stimuli and apparatus
These were the same as those in Experiment 1.
Procedure
Experiment 2 consisted of 52 trials divided into
four blocks, one block for each combination of
word type (concrete, abstract) and concurrent task
(random tapping, simple tapping). Two participants were randomly assigned to each of the 24
possible orders of presentation of the four blocks.
Before the experimental task, participants were
instructed on how to perform the two kinds of
tapping. For the simple-tapping condition, they
were asked to click the mouse button at a steady
pace of about two clicks per second. For the
random-tapping condition, participants were
instructed to click the mouse button at variable
intervals, trying to generate the most unpredictable
and haphazard sequence they could produce.
Each trial began with a message indicating the
kind of tapping (simple or random) that was
Table 3. Percentage of correct responses, order errors, and item errors
in Experiment 2
Word type
Concrete
Abstract
Concurrent
task
Correct
responses
Order
errors
Item
errors
Random
tapping
Simple tapping
Random
tapping
Simple tapping
67.02
18.73
18.14
74.68
61.67
14.42
21.01
13.17
22.37
69.36
16.67
17.37
Note: Standard deviations in parentheses.
required in that trial, which remained on the
screen until the first participant’s click of the
tapping sequence. After two seconds, a plus sign
appeared on the computer screen for one second
followed by a five-word list. Words were presented
at a rate of one word every two seconds with participants being instructed to perform the tapping task
throughout the entire list presentation (but not
during recall). Other aspects of the procedure
were the same as those in Experiment 1.
Results
As in Experiment 1, participants’ responses in the
memory task were classified into correct responses,
order errors, and item errors (Table 3). Table 4
Table 4. Statistical results for the 2 (word type) × 2 (concurrent
task) within-subjects ANOVA on the percentage of correct responses,
order errors, and item errors in Experiment 2
Effect
Correct responses
Word type
Concurrent task
Word Type × Task
Order errors
Word type
Concurrent task
Word Type × Task
Item errors
Word type
Concurrent task
Word Type × Task
F
df
MSE
p
h2p
17.839 1, 47 76.619 ,.001 .275
44.939 1, 47 62.942 ,.001 .489
0.000 1, 47 37.555
.986 .000
5.314 1, 47 46.447
.026 .102
19.818 1, 47 45.333 ,.001 .297
0.000 1, 47 33.791
.992 .000
22.297 1, 47 38.246 ,.001 .322
37.167 1, 47 32.086 ,.001 .442
0.001 1, 47 17.355
.979 .000
Note: ANOVA = analysis of variance.
THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014
7
Downloaded by [the Bodleian Libraries of the University of Oxford] at 08:54 11 November 2014
CAMPOY ET AL.
shows the statistical results of the ANOVAs
detailed below.
Percentages of correct responses were submitted
to a 2 × 2 within-subjects ANOVA with word type
(concrete, abstract) and concurrent task (simple
tapping, random tapping) as factors. There was a
main effect of word type, showing that concrete
words were better recalled than abstract words
(concreteness effect = 5.34%). The main effect of
concurrent task was also significant, revealing
better memory performance in the simple-tapping
condition than in the random-tapping condition
(tapping effect = 7.68%). Importantly, however,
there was no interaction between word type and
concurrent task, showing that the concreteness
effect did not differ between tapping conditions.
An ANOVA on the percentages of order errors
revealed significant main effects of both word type
(more order errors for abstract words) and concurrent task (more order errors with random tapping).
In turn, an ANOVA on the percentages of item
errors also showed significant main effects of
word type (more item errors with abstract words)
and concurrent task (more item errors with
random tapping). In both cases, however, there
was no significant interaction between the two
factors.
Additional analyses were conducted to examine
participants’ performance on the secondary task.
Specifically, our interest focused on the level of randomness in the tapping patterns generated by the
participants. We used the variability of the
between-tap intervals as an index of randomness,
with higher variability being assumed to reflect
more random patterns. Interval variability in a
trial was quantified as the average proportional
difference between all intervals generated by the
participant in that trial (for a detailed description
of these calculations, see Heath, 2006, Formulae
1, 2, and 4). Intervals generated before the onset
of the first to-be-remembered word were not considered. Values of interval variability were .40 for
both concrete and abstract trials in the randomtapping condition and .17 for both concrete and
abstract trials in the simple-tapping condition. A
2 × 2 within-subjects ANOVA with word type
and tapping type as factors showed a main effect
8
of tapping type, F(1, 47) = 330.183, MSE =
85.386, p , .001, h2p = .875, revealing greater
interval variability in the random-tapping condition
than in the simple-tapping condition. Neither the
main effect of word type nor the interaction was
significant (both Fs , 1).
Discussion
Results showed that introducing a concurrent
attention-demanding task (random tapping)
affected memory performance negatively, with
poorer immediate serial recall being a consequence
of both order errors and item errors. Contrary to
the strategic hypothesis, however, limiting available attentional resources by a concurrent task
had parallel negative effects on the recall of concrete and abstract words, so that the concreteness
effect in the random-tapping condition was equivalent to that obtained in a control, simple-tapping
condition. In turn, the level of randomness in the
tapping sequences (estimated from the variability
of the generated intervals) was numerically identical in concrete and abstract trials. This seems to
rule out the possibility that better memory performance for concrete words was a consequence
of neglecting the tapping task during the presentation of concrete lists. These results are clearly
inconsistent with the idea that the concreteness
effect in STM is a consequence of controlled
attention-dependent processes, as suggested by
the strategic hypothesis.
One additional aspect of the results that deserves
consideration is the fact that, in contrast to
Experiment 1, there was a significant effect of
word concreteness in terms of order, as occasionally
found in previous studies (Acheson et al., 2010,
Experiment 1; Allen & Hulme, 2006,
Experiment 1). An inspection of Table 4,
however, reveals that the size of the concreteness
effect (h2p ) in terms of item errors was three times
larger than the effect in terms of order errors. It
seems safe, therefore, to conclude that the concreteness effect was mostly a consequence of differences
in item recall. In any case, it is important to note
that percentage of order and item errors represents
only a rough, indirect estimation of order and item
THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014
AUTOMATIC SEMANTIC ENCODING
memory, respectively, and should be interpreted
with caution.
Downloaded by [the Bodleian Libraries of the University of Oxford] at 08:54 11 November 2014
EXPERIMENT 3
At odds with the strategic hypothesis, Experiment
2 found that limiting available attention resources
by a concurrent task did not eliminate or reduce
the concreteness effect. Experiment 3 aimed to
further investigate the consequences of introducing
concurrent attention-demanding tasks by a procedure that was presumed to maximize the opportunity for the concurrent task to detectably affect
the concreteness effect.
Main differences with respect to Experiment 2
were twofold. On the one hand, the closed sets of
stimuli used in the previous experiment were
replaced by open sets of words, so that each experimental stimulus appeared only once during
the experiment. As found previously in this and
other studies (e.g., Walker & Hulme, 1999), the
concreteness effect in STM seems to be mainly
a consequence of differences in item recall.
Consequently, a procedure with open sets of
stimuli was expected to be more sensitive because
performance would depend on item memory to a
greater extent. On the other hand, we introduced
concurrent tasks involving visuospatial interference.
We employed two different concurrent tasks, one
predominantly visual and the other predominantly
spatial. In one condition, three irregular polygons
were presented simultaneously with the auditory
presentation of each to-be-remembered word, and
participants had to determine which polygon was
different from the other two by pressing the corresponding key (see Figure 1). This task was assumed
to involve high levels of visual interference (cf.
Logie, 1995), although spatial interference was
probably present because of the fact that there
Figure 1. Schematic representation of the experimental procedure in the polygon condition of Experiment 3. Participants responded to each
polygon triad, though only the first response is depicted as an example (represented response is correct). Depicted polygons correspond to
stimuli actually used in the experiment.
THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014
9
Downloaded by [the Bodleian Libraries of the University of Oxford] at 08:54 11 November 2014
CAMPOY ET AL.
Figure 2. Schematic representation of the experimental procedure in the arrow condition of Experiment 3. Participants responded to each arrow
triad, though only the first response is depicted (represented example corresponds to an incorrect response).
was a spatial stimuli–response mapping (see below).
In another condition, three arrows were presented
with each word, and participants had to reproduce
the sequence by pressing the corresponding arrow
keys (see Figure 2). Interference in this condition
was assumed to be mainly spatial, though the
visual presentation of the arrows probably generated considerable levels of visual interference.
Apart from limiting the availability of attentional
resources, the fact that these concurrent tasks
involved high levels of visuospatial interference
was expected to hinder a potentially important
semantic elaborative strategy: that based on the
generation of mental images and visual scenes
(Paivio, 1986). Because concrete words are more
imageable than abstract words, it seems reasonable
to presume that such strategy would benefit the
recall of concrete words to a greater extent.
Evidence for the fact that the presentation of
visual inputs interferes with the generation of
mental images has been found in a wide range of
10
studies employing different approaches (Baddeley
& Andrade, 2000; Dean, Dewhurst, Morris, &
Whittaker, 2005; Logie, 1986; Quinn &
McConnell, 2006). According to Borst, Niven,
and Logie (2012), the presentation of visual
inputs interferes with the generation of mental
images because the mechanisms underlying visual
mental imagery overlap to a great extent to those
involved in visual perception. Congruent with this
notion, visual mental imagery and visual perception
have been found to recruit overlapping sensory
regions of the brain (Kosslyn et al., 1993).
Concurrent task was manipulated between participants to guarantee sufficient number of observations per condition without increasing the
duration of the task and the complexity of the
design. Participants’ performance in the two concurrent-task conditions was compared to that in a
control condition without concurrent task. The
prediction from the strategic hypothesis was that
the concreteness effect would be eliminated or
THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014
AUTOMATIC SEMANTIC ENCODING
reduced in the concurrent-tasks conditions. No
specific prediction was made about differences
between the predominantly visual and the spatial
concurrent tasks.
Downloaded by [the Bodleian Libraries of the University of Oxford] at 08:54 11 November 2014
Method
Participants
One hundred and seventeen undergraduate students from the University of Murcia participated
in the experiment. Thirty-nine participants were
randomly assigned to each of the two concurrenttask conditions (polygons and arrows). The
remaining 39 participants were assigned to a
control group with no concurrent task.
Stimuli and apparatus
We selected two sets of Spanish words from the
LEXESP database (Sebastián et al., 2000), one
comprising 70 concrete, high-imageability words,
and the other comprising 70 abstract, low-imageability words. All of the words were trisyllabic
nouns with the stress on the penultimate syllable.
The two word sets were globally matched for
number of phonemes and individually matched
for word frequency and familiarity (see
Appendix). An additional group of 30 trisyllabic
nouns with similar values of frequency and familiarity were chosen for practice trials.
All the words were recorded by a female speaker
in a neutral tone of voice and were segmented into
individual sound files using audio editing software.
The experiment was controlled by experimental
software written in E-Prime (Schneider et al.,
2002). Additionally, we used custom-built software
written in Visual Basic (Microsoft Co., Redmond,
WA, USA) to generate word lists for each participant prior to the experimental session. To minimize the phonological similarity of the words
selected for a given list, no words within a list
were allowed to share a syllable in the same position
or to start with the same phoneme. Apart from this
restriction, the programme created the lists by
selecting words from the pertinent set at random
without replacement. A headset with microphone
was used to present stimuli and record participants’
responses.
For the polygons condition, we constructed 168
pairs of eight-sided irregular polygons by using a
program written in Matlab (The Mathworks,
Inc., Natick, MA, USA) by Collin and
McMullen (2002). This programme was designed
to generate families of polygons, with the
members of a family being similar to each other
(see Collin & McMullen, 2002, for an operational
definition of family similarity). With the exception
of the number of sides, we used the default values
for all the customizable parameters of the programme to generate 168 families of polygons. We
then took two instances from each family to
obtain the 168 pairs.
Procedure
The experiment consisted of 28 trials, 14 trials for
each type of word. The experimental procedure in
the polygons condition is depicted in Figure 1.
Each trial began with the presentation of a row of
dashes, indicating that a key had to be pressed to
initiate the trial. After 1.5 seconds following the
key press, three polygons were presented in a horizontal row, and participants had to indicate which
polygon differed from the other two by pressing
the pertinent key. The polygon triad consisted of
the two polygons composing one of the 168 pairs
of similar polygons (see above), with one of the
polygons of the pair appearing twice. Participants
were instructed to press the key 1, 2, or 3 on the
numeric keypad according to the locus of the differing polygon (key 1, 2, and 3 for left, middle, and
right positions, respectively). The polygon triad
remained on the screen for 2 s regardless of the participant’s response. After the first polygon triad, a
five-word list was presented auditorily at a rate of
one word every two seconds, with each word
being accompanied by the visual presentation of a
new polygon triad. The locus of the differing polygons within the triads was determined at random
with the constraint that, throughout each trial,
these polygons appeared twice in each of the
three possible positions. After the last stimulus of
the list, a row of five question marks was presented
to prompt participants for spoken recall of the five
words in serial order. They were asked to substitute
the word espacio (blank) for any word they could not
THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014
11
CAMPOY ET AL.
Table 5. Percentage of correct responses, order errors, and item errors in Experiment 3
Word type
Concurrent task
Concrete
Polygons
Arrows
Control
Polygons
Arrows
Control
Downloaded by [the Bodleian Libraries of the University of Oxford] at 08:54 11 November 2014
Abstract
Correct responses
Order errors
Item errors
59.01 (13.09)
61.06 (14.02)
73.55 (14.81)
51.79 (14.97)
51.61 (16.34)
64.95 (14.98)
19.23 (9.56)
17.68 (8.75)
9.73 (10.09)
19.91 (10.77)
20.80 (13.33)
11.94 (8.96)
27.47 (10.99)
26.37 (12.18)
19.08 (10.52)
36.30 (13.28)
35.71 (14.09)
26.81 (12.10)
Note: Standard deviations in parentheses.
recall. There was a 16-s time limit to complete
recall. The procedure in the arrows condition was
the same as that in the polygons condition, but
polygon triads were replaced by three arrows pointing to different directions (see Figure 2). For each
arrow presentation, participants had to press the
corresponding arrow keys on the keyboard, starting
from the arrow presented on the left of the screen.
In the control condition, no polygons or arrows
were presented, with other aspects of the procedure
being the same. In the three conditions, six practice
trials preceded experimental trials.
Results
Participants’ responses in the memory task were
classified into correct responses, order errors, and
item errors as in previous experiments (Table 5).
Table 6 presents the statistical results of the
ANOVAs described below.
Percentages of correct responses were submitted
to a 2 × 3 mixed ANOVA with word type (concrete, abstract) as the within-subject factor and
concurrent task (polygons, arrows, and control) as
the between-subject factor. There was a main
effect of word type, with concrete words being
better recalled than abstract words (concreteness
effect = 8.43%). There was also a main effect of
concurrent task, revealing that the percentage of
correct responses differed between groups. Post
hoc Tukey HSD tests (MSE = 381.48, df = 114)
revealed better memory performance in the control
group that in the other two groups (both
ps , .001; polygon effect = 13.85%; arrow effect =
12.91%), whereas there was no difference between
the polygon group and the arrow group
Table 6. Statistical results for the 2 (word type) × 3 (concurrent task) mixed ANOVA on the percentage of correct responses, order errors, and
item errors in Experiment 3
F
df
MSE
p
h2p
78.522
12.245
0.469
1, 114
1, 114
2, 114
52.884
381.475
52.884
,.001
,.001
.627
.408
.177
.008
5.534
11.119
0.693
1, 114
1, 114
2, 114
42.638
171.949
42.638
.020
,.001
.502
.046
.163
.012
131.664
7.107
0.399
1, 114
1, 114
2, 114
33.111
267.268
33.111
,.001
.001
.672
.536
.111
.007
Effect
Correct responses
Word type
Concurrent task
Word Type × Task
Order errors
Word type
Concurrent task
Word Type × Task
Item errors
Word type
Concurrent task
Word Type × Task
Note: ANOVA = analysis of variance.
12
THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014
Downloaded by [the Bodleian Libraries of the University of Oxford] at 08:54 11 November 2014
AUTOMATIC SEMANTIC ENCODING
(p = .952). Importantly, there was no interaction
between word type and concurrent tasks, revealing
that the concreteness effect did not differ across
groups.
An ANOVA on the percentages of order errors
showed a significant main effect of word type (more
order error with abstract words). The main effect of
concurrent task was also significant, with post hoc
Tukey HSD tests (MSE = 171.95, df = 114)
revealing significant differences between the
control group and the other two groups (fewer
order errors in the control groups; both
ps , .001). There was no difference between the
polygons group and the arrows group (p = .987).
The interaction between word type and concurrent
task was not significant. Similarly, an ANOVA on
the percentages of item errors revealed a main effect
of both word type (more item errors with abstract
words) and concurrent task. Post hoc Tukey
HDS tests (MSE = 267.27, df = 114) showed significant differences between the control group and
the other two groups (fewer item errors in the
control group, both ps , .01), with no difference
between the polygons group and the arrows group
(p = .945). Again, there was no interaction
between word type and concurrent task.
In order to analyse performance in the concurrent
tasks, we calculated the proportion of correct
responses and the mean reaction time (RT) on concrete and abstract trials for each participant.
Responses generated prior to presentation of the
first word in the list (responses to the first polygon/
arrow triad in each trial) were not considered.
Mean proportion of correct responses was .72 in
both the polygon and arrow groups and for both concrete-word and abstract-word trials. As expected
from this numerical coincidence, a 2 × 2 ANOVA
on the proportion of correct responses with word
type and concurrent task as factors showed no significant main effects or interaction (all Fs , 1). In
turn, an ANOVA on the mean RTs revealed a
main effect of concurrent task (longer RTs in the
arrow group), F(1, 76) = 33.864, MSE =
18,575.557, p , .001, h2p = .308. However, neither
the main effect of word type nor the interaction
was significant (both Fs , 1). Mean RT for concrete
and abstract trials were, respectively, 1300 ms and
1290 ms in the polygon group, and 1378 ms and
1375 ms in the arrow group.
Discussion
Results were equivalent to those in Experiment
2. In comparison with the control condition, the
introduction of concurrent demanding tasks had a
detrimental effect on memory performance, involving more item errors and order errors.
Importantly, however, this detrimental effect of
the concurrent tasks on recall was equivalent for
concrete and abstract lists, so that the concreteness
effect on the conditions with concurrent tasks was
comparable to that in the control condition.
Performance in the concurrent tasks was equivalent
for concrete and abstract trials, ruling out the possibility that better recall of concrete words was due to
selective neglect of the concurrent task during the
presentation of concrete lists. As in Experiment
2, therefore, results in this experiment were incompatible with the strategic hypothesis and rather
support the idea that semantic encoding in standard STM tasks relies on automatic mechanisms.
Experiment 3 again showed reliable concreteness effects in terms of both item and order
errors. However, an inspection of effect sizes in
Table 6 revealed that the concreteness effect in
terms of item errors was more than 11 times the
size of the effect in terms of order errors, consistent
with the idea that lexical–semantic factors in verbal
STM tasks principally affect item memory rather
than order memory (Majerus, 2009).
GENERAL DISCUSSION
The concreteness effect in STM is broadly assumed
to be a consequence of semantic traces being richer
and more distinctive for concrete than for abstract
words (Walker & Hulme, 1999). From this point
of view, the concreteness effect can be taken as an
index of the degree to which the to-be-remembered
words are semantically encoded in STM. We
adopted this approach to evaluate the strategic
hypothesis: the idea that the participation of
semantic codes in standard STM tasks (those
THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014
13
Downloaded by [the Bodleian Libraries of the University of Oxford] at 08:54 11 November 2014
CAMPOY ET AL.
involving the immediate serial recall of lists of unrelated words) relies on controlled, strategic mechanisms of semantic retrieval and elaborative encoding.
Since controlled semantic retrieval and elaborative encoding are most probably time dependent,
the strategic hypothesis would predict larger
semantic effects when to-be-remembered words
were presented at slower presentation rates.
Congruently with this prediction, Experiment 1
showed a larger concreteness effect with a presentation rate of one word every two seconds in comparison with a rate of one word per second. The
concreteness effect in this experiment was specifically a consequence of differences between concrete
and abstract words in terms of item errors, with the
presentation rate affecting exclusively the percentage of item errors with concrete words. Neither
order errors nor memory performance for abstract
words was modulated by the presentation rate.
Although the results of Experiment 1 were
compatible with the strategic hypothesis, they
only constitute indirect evidence, with alternative
interpretations being also possible (see below).
Experiments 2 and 3 provided a more direct test
of the strategic hypothesis by introducing concurrent attention-demanding tasks: a random time
interval generation task in Experiment 2 and two
different visuospatial tasks in Experiment 3. In
contrast with the random time interval generation
task, which was originally conceived to demand
executive resources with no verbal or visuospatial
interference (Vandierendonck et al., 1998), concurrent tasks in Experiment 3 involved extensive
visuospatial interference. We expected that this
visuospatial interference would hinder potentially
relevant semantic strategies based on the generation
of mental images or scenes. Another important
difference between Experiments 2 and 3 is that
Experiment 3 included open sets of stimuli (with
each to-be-remembered word appearing only once
throughout the experiment), instead of the closed
sets used in Experiment 2. We assumed that
memory performance would rely on item memory
to a greater extent with open sets of stimuli, so
that the procedure would be more sensitive to variations in the concreteness effect. Despite these
differences, Experiments 2 and 3 yielded a
14
qualitatively identical pattern. Reducing available
attentional resources by concurrent, attentiondemanding tasks resulted in a marked decrement
in memory performance, expressed in worse item
and order recall. However, the concreteness effect
in conditions with concurrent demanding tasks
was equivalent to that found in control conditions.
These results constitute strong evidence against the
strategy hypothesis and, instead, support the idea
that the concreteness effect in standard STM
tasks emerges from automatic encoding of semantic
information.
If semantic encoding in STM tasks does not
depend on a controlled, time-dependent mechanism, an alternative interpretation is required for
the effect of presentation rate in Experiment
1. One possibility is to evoke the levels of processing (LoP) principle, whereby the more deeply
and elaborately the stimulus is processed, the
better it is retained (Craik & Lockhart, 1972). It
might be argued that a slow presentation rate
could favour deeper (semantic) processing during
word presentation, which would benefit especially
the subsequent recall of concrete words. However,
previous research has shown that the LoP only
has an effect on immediate recall when the
memory task includes a highly interfering secondary task that entails massive displacement of information from STM to long-term memory (Loaiza,
McCabe, Youngblood, Rose, & Myerson, 2011;
Rose & Craik, 2012; Rose, Myerson, Roediger,
& Hale, 2010). An explanation based on the
LoP, thus, seems implausible. Alternatively, if we
assume that semantic encoding is automatic but
involves spreading activation (Collins & Loftus,
1975) throughout the semantic network, then
more time will lead to broader and deeper activation, better encoding, and better recall. In the
case of phonological encoding, the associated
network is much less rich and hence gains less
from the added time. Otherwise, within the framework of the multicomponent working memory
model (Baddeley & Hitch, 1974), it has been
recently suggested that a slower presentation rate
might promote the participation of the episodic
buffer as the main storage device for the immediate
recall of verbal information (Barrouillet, Plancher,
THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014
Downloaded by [the Bodleian Libraries of the University of Oxford] at 08:54 11 November 2014
AUTOMATIC SEMANTIC ENCODING
Guida, & Camos, 2013). The episodic buffer is
assumed to store multimodal representations that
combine features of different nature (for example,
verbal, visual, and semantic) in a multidimensional
code, integrating information from diverse modalities and memory systems (Baddeley, 2000).
Constructing this kind of multidimensional representations requires taking and binding information from different sources into a single
representation, with these processes probably benefiting from a slower presentation rate. From this
perspective, the effect of presentation rate in
Experiment 1 might be due to the greater involvement of the episodic buffer with slow presentations,
with concrete words benefiting from richer and
more distinctive multidimensional representation
in this buffer.
An additional aspect that deserves consideration
is the global decrement of memory performance
yielded by the introduction of attention-demanding tasks in Experiments 2 and 3. In our opinion,
this result may be a consequence, at least in part,
of the effect of reducing available attentional
resources on the maintenance mechanisms in
STM. At present, maintenance of verbal information in the short term is usually assumed to
rely on two different mechanisms: articulatory
rehearsal and attentional refreshing (Camos,
Lagner, & Barrouillet, 2009; Mora & Camos,
2013; Oberauer & Lewandowsky, 2008, 2013;
Raye, Johnson, Mitchell, Greene, & Johnson,
2007). The reason why introducing attentiondemanding concurrent tasks could have interfered
with the attention-based refreshing mechanism
seems obvious because this mechanism is assumed
to rely on domain-general attentional resources,
with the same limited resources being required for
both refreshing STM traces and performing the
concurrent tasks (Barrouillet, Bernardin, &
Camos, 2004; Oberauer & Lewandowsky, 2008,
2013). Using concurrent tasks to temporally
hinder attentional refreshing constitutes the basic
procedure of the studies on the time-based
resource-sharing model of working memory,
which have repeatedly shown how memory performance decreases as the concurrent task distracts
attention for a greater proportion of time
(Barrouillet et al., 2004; Barrouillet, Bernardin,
Portrat, Vergauwe, & Camos, 2007). Regarding
articulatory rehearsal, it has been generally
assumed that this maintenance mechanism operates with no extensive attentional demand, but
that attention is necessary during the very initial
phase in which the rehearsal loop is set up
(Naveh-Benjamin & Jonides, 1984). Therefore,
limiting attentional resources could have impaired
memory performance by both restricting the participation of the attentional refreshing mechanism
and disrupting the construction of the rehearsal
loops.
Our final considerations concern the question
of what might be the nature of the semantic processing underlying the concreteness effect in
STM. One possibility is that this effect was a consequence of participants encoding and actively
maintaining semantic representations. Initial proposals about the contribution of semantic maintenance in verbal STM tasks emerged from
neuropsychological patients showing what was
claimed to be a specific deficit in the short-term
retention of semantic representations (Martin,
Shelton, & Yaffee, 1994). Subsequent research
suggests that STM performance of these patients
is a consequence of the impairment of semantic
control mechanisms rather than a specific semantic
STM deficit (Hoffman et al., 2011). Either way, it
has been alleged that memory performance of
these patients in verbal STM task is not necessarily related to the active maintenance of semantic
representations in the short term (Shivde &
Anderson, 2011). Alternative accounts include,
for example, a poorer use of long-term memory
at recall, resulting in less efficient reconstruction
of degraded phonological traces (Shivde &
Anderson, 2011). From a different approach,
Shivde and Anderson (2011) found evidence of
semantic maintenance by using what they called
the concurrent probe paradigm. This procedure
included two apparently independent tasks that
participants had to perform simultaneously. On
the one hand, participants were asked to keep in
mind the meaning of a target word in order to
indicate whether this meaning matched the
meaning of a probe word that was presented
THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014
15
Downloaded by [the Bodleian Libraries of the University of Oxford] at 08:54 11 November 2014
CAMPOY ET AL.
after a retention interval. Concurrently, participants performed a lexical decision task, which
could occur either during the retention interval
or after the participants’ response to the memory
probe. Analysis of the reaction times in the
lexical decision task revealed slower responses for
words semantically related to the target item that
they were asked to maintain. Other aspects being
equal, this semantic relatedness effect vanished if
participants had already provided a response to
the probe word. There was also no effect in equivalent conditions in which the relevant dimension
in the memory task was the phonological or
visual (word form) characteristics of the target
and probe words rather than their meaning. The
effect also disappeared when participants were
instructed to focus on the lexical decision task
and stop maintaining in mind the meaning of
the target word. These findings were interpreted
as a proof that the relatedness effect was actually
a consequence of active semantic maintenance.
In principle, it could be alleged that a similar
form of active semantic maintenance might
underlie the concreteness effect found in the
present study. However, it is important to note
that, unlike our study, both the instructions provided to the participants and the nature of the
memory task itself oriented participants to
engage in semantic maintenance. This contrasts
with what occurs in standard verbal STM procedures, in which it has been broadly demonstrated that participants mainly rely on the
maintenance of phonological traces. The same
argument can be raised with regard to a number
of neuroimaging studies that seem to identify the
neural basis of short-term semantic maintenance:
a frontotemporal circuit in which prefrontal areas
(mainly, the LIPFC) would be in charge of maintaining the activation of neural semantic representations in the temporal cortex (Fiebach, Friederici,
Smith, & Swinney, 2007; Shivde & Thompson,
2004). Again, memory tasks in these neuroimaging studies were designed to boost semantic
encoding and maintenance. The fact that there is
no evidence of active semantic maintenance with
a standard immediate serial recall procedure casts
doubt about the possibility that such active
16
maintenance mechanisms underlie semantic
effects in this kind of tasks, with alternative
accounts being possible. It could be alleged, for
example, that semantic effects in standard tasks
are a consequence of the initial semantic activation
generated by the presentation of a to-be-remembered words facilitating the reconstruction of
degraded phonological traces. This facilitating
effect could be seen as a form of semantic
priming (Shivde & Anderson, 2011) and would
operate at recall but also during the retention
interval, contributing to the reactivation and maintenance of phonological traces in this interval. In
its turn, reactivation of phonological traces
during the retention interval might involve the
subsidiary automatic reactivation of semantic representations leading to a snowball effect, with
semantic activation facilitating the reactivation of
phonological traces and reactivation of phonological traces feeding back the activation of semantic
representation.
We began with a simple question, whether the
semantic contribution to immediate serial verbal
recall depended on the application of a controlled
semantic strategy. Our results suggest that this is
not the case, and, as such, they resemble findings
with more complex sentence stimuli. Baddeley,
Hitch, and Allen (2009) studied the effects of concurrent disruption of the visuospatial, phonological, and executive subcomponents of working
memory on the capacity to take advantage of sentence structure in immediate serial recall, comparing retention of constrained sentences with that of
the equivalent words in random order. They found
clear effects of visuospatial and articulatory suppression, which were magnified when combined
with an attentional load. Each concurrent task
had a greater or lesser impact on overall performance, but this did not interact with the advantage
gained from sentence form. This is also consistent
with the conclusions drawn by Caplan and Waters
(1999) who present substantial evidence that the
use of grammatical structure in language processing is automatic and does not require additional
executive processes.
To summarize, the present study shows that
semantic encoding in standard verbal STM
THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014
Downloaded by [the Bodleian Libraries of the University of Oxford] at 08:54 11 November 2014
AUTOMATIC SEMANTIC ENCODING
tasks (immediate serial recall of unrelated words)
does not depend on the participation of controlled semantic strategies. It is important to
note, however, that this does not mean that participants cannot implement this kind of strategy
in certain circumstances, as suggested by the
fact that the standard phonological effect in
STM is abolished when participants are explicitly
instructed to use a semantic strategy (Campoy &
Baddeley, 2008).
Original manuscript received 14 May 2013
Accepted revision received 18 August 2014
REFERENCES
Acheson, D. J., Postle, B. R., & MacDonald, M. C.
(2010). The interaction of concreteness and phonological similarity in verbal working memory. Journal of
Experimental Psychology: Learning, Memory, and
Cognition, 36, 17–36.
Allen, R., & Hulme, C. (2006). Speech and language
processing mechanisms in verbal serial recall. Journal
of Memory and Language, 55, 64–88.
Baddeley, A. D. (1966). Short-term memory for word
sequences as a function of acoustic, semantic and
formal similarity. Quarterly Journal of Experimental
Psychology, 18, 362–365.
Baddeley, A. D. (2000). The episodic buffer: A new
component of working memory?. Trends in
Cognitive Science, 4, 417–423.
Baddeley, A. D. (2012). Working memory: Theories,
models, and controversies. Annual Review of
Psychology, 63, 1–29.
Baddeley, A. D., & Andrade, J. (2000). Working
memory and the vividness of imagery. Journal of
Experimental Psychology: General, 129, 126–145.
Baddeley, A. D., & Ecob, J. R. (1970). Simultaneous
acoustic and semantic coding in short-term
memory. Nature, 227, 288–289.
Baddeley, A. D., & Hitch, G. J. (1974). Working
memory. In G. H. Bower (Ed.), The psychology of
learning and motivation: Advances in research and
theory, Vol. 8 (pp. 47–90). New York: Academic.
Baddeley, A. D., Hitch, G. J., & Allen, R. J. (2009).
Working memory and binding in sentence recall.
Journal of Memory and Language, 61, 438–456.
Baddeley, A. D., & Levy, B. A. (1971). Semantic coding
and short-term memory. Journal of Experimental
Psychology, 89, 132–136.
Baddeley, A. D., Thomson, N., & Buchanan, M. (1975).
Word length and the structure of short-term memory.
Journal of Verbal Learning and Verbal Behavior, 14,
575–589.
Badre, D., Poldrack, R. A., Pare-Blagoev, E. J., Insler, R.
Z., & Wagner, A. D. (2005). Dissociable controlled
retrieval and generalized selection mechanisms in
ventrolateral prefrontal cortex. Neuron, 47, 907–918.
Badre, D., & Wagner, A. D. (2002). Semantic retrieval,
mnemonic control, and prefrontal cortex. Behavioral
and Cognitive Neuroscience Reviews, 1, 206–218.
Barrouillet, P., Bernardin, S., & Camos, V. (2004). Time
constraints and resource sharing in adults’ working
memory spans. Journal of Experimental Psychology:
General, 133, 83–100.
Barrouillet, P., Bernardin, S., Portrat, S., Vergauwe, E.,
& Camos, V. (2007). Time and cognitive load in
working memory. Journal of Experimental Psychology:
Learning, Memory, and Cognition, 33, 570–585.
Barrouillet, P., Plancher, G., Guida, A., & Camos, V.
(2013). Forgetting at short term: When do eventbased interference and temporal factors have an
effect? Acta Psychologica, 142, 155–167.
Borst, G., Niven, E., & Logie, R. H. (2012). Visual
mental image generation does not overlap with
visual short-term memory: A dual-task interference
study. Memory and Cognition, 40, 360–372.
Burgess, N., & Hitch, G. J. (1999). Memory for
serial order: A network model of the phonological
loop and its timing. Psychological Review, 106, 551–
581.
Camos, V., Lagner, P., & Barrouillet, P. (2009). Two
maintenance mechanisms of verbal information in
working memory. Journal of Memory and Language,
61, 457–469.
Campoy, G., & Baddeley, A. (2008). Phonological and
semantic strategies in immediate serial recall.
Memory, 16, 329–340.
Caplan, D., & Waters, G. S. (1999). Verbal working
memory and sentence comprehension. Behavioral
and Brain Sciences, 22, 77–94.
Collin, C. A., & McMullen, P. (2002). Using Matlab to
generate families of similar Attneave shapes.
Behaviour Research Methods, Instruments, and
Computers, 34, 55–68.
Collins, A. M., & Loftus, E. F. (1975). A spreadingactivation
theory
of
semantic
processing.
Psychological Review, 82, 407–428.
THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014
17
Downloaded by [the Bodleian Libraries of the University of Oxford] at 08:54 11 November 2014
CAMPOY ET AL.
Conrad, R., & Hull, A. J. (1964). Information, acoustic
confusion, and memory span. British Journal of
Psychology, 55, 429–432.
Cowan, N. (2005). Working memory capacity. New York,
NY: Psychology Press.
Craik, F. I. M., & Lockhart, R. S. (1972). Levels of
processing: A framework for memory research.
Journal of Verbal Learning and Verbal Behavior, 11,
671–684.
Dean, G. M., Dewhurst, S. A., Morris, P. E., &
Whittaker, A. (2005). Selective interference with
the use of visual images in the symbolic distance paradigm. Journal of Experimental Psychology: Learning,
Memory, and Cognition, 31, 1043–1068.
Fiebach, C. J., Friederici, A. D., Smith, E. E., &
Swinney, D. (2007). Lateral inferotemporal cortex
maintains conceptual–semantic representations in
verbal working memory. Journal of Cognitive
Neuroscience, 19, 2035–2049.
Gold, B. T., Balota, D. A., Jones, S. J., Powell, D. K.,
Smith, C. D., & Andersen, A. H. (2006).
Dissociation of automatic and strategic lexicalsemantics: Functional magnetic resonance imaging
evidence for differing roles of multiple frontotemporal
regions. Journal of Neuroscience, 26, 6523–6532.
Haarmann, H. J., & Usher, M. (2001). Maintenance of
semantic information in capacity-limited item short
term memory. Psychonomic Bulletin and Review, 8,
568–578.
Heath, J. P. (2006). Quantifying temporal variability in
population abundances. Oikos, 115, 573–581.
Hoffman, P., Jefferies, E., & Ralph, M. A. (2011).
Explaining semantic short-term memory deficits:
Evidence for the critical role of semantic control.
Neuropsychologia, 49, 368–381.
Hulme, C. Maughan, S., & Brown, G. D. A. (1991).
Memory for familiar and unfamiliar words:
Evidence for a long-term memory contribution to
short-term memory span. Journal of Memory and
Language, 30, 685–701.
Hulme, C., Roodenrys, S., Schweickert, R., Brown,
G. D. A., Martin, S., & Stuart, G. (1997). Word frequency effects on short-term memory tasks: Evidence
for a redintegration proces in immediate serial recall.
Journal of Experimental Psychology: Learning,
Memory, and Cognition, 23, 1217–1232.
Kosslyn, S. M., Alpert, N. M., Thompson, W. L.,
Maljkovic, V., Weise, S. B., Chabris, C. F., …
Buonnano, F. S. (1993). Visual mental imagery activates topographically organized visual cortex: PET
18
investigations. Journal of Cognitive Neuroscience, 5,
263–287.
Loaiza, V., McCabe, D., Youngblood, J., Rose, N. S., &
Myerson, J. (2011). The influence of levels of
processing on recall from working memory and
delayed recall tasks. Journal of Experimental
Psychology: Learning, Memory, and Cognition, 37,
1258–1263.
Logie, R. H. (1986). Visuo-spatial processing in working
memory. The Quarterly Journal of Experimental
Psychology, 38A, 229–247.
Logie, R. H. (1995). Visuo-spatial working memory.
Hove, UK: Lawrence Erlbaum.
Logie, R. H., Della Sala, S., Laiacona, M., Chalmers, P.,
& Wynn, V. (1996). Group aggregates and individual
reliability: The case of verbal short-term memory.
Memory and Cognition, 24, 305–321.
Majerus, S. (2009). Verbal short-term memory and temporary activation of language representations: The
importance of distinguishing item and order information. In A. S. C. Thorn & M. P. A. Page
(Eds.), Interactions between short-term and long-term
memory in the verbal domain (pp. 244–276). Hove,
UK: Psychology Press.
Martin, R. C. (2005). Components of short-term
memory and their relation to language processing.
Evidence from neuropsychology and neuroimaging.
Current Directions in Psychological Science, 14, 204–208.
Martin, R. C., Shelton, J., & Yaffee, L. (1994).
Language processing and working memory:
Neuropsychological evidence for separate phonological and semantic capacities. Journal of Memory and
Language, 33, 83–111.
Mora, G., & Camos, V. (2013). Two systems of maintenance in verbal working memory: Evidence from
the word length effect. PLoS One, 8, Article no.
e70026.
Murdock, B. B. (1976). Item and order information in
short-term serial memory. Journal of Experimental
Psychology: General, 105, 191–216.
Nairne, J. S., & Kelley, M. R. (2004). Separating item
and order information through process dissociation.
Journal of Memory and Language, 50, 113–133.
Naveh-Benjamin, M., & Jonides, J. (1984). Maintenance
rehearsal: A two-component analysis. Journal of
Experimental Psychology: Learning, Memory and
Cognition, 10, 369–385.
Oberauer, K., & Lewandowsky, S. (2008). Forgetting in
immediate serial recall: Decay, temporal distinctiveness,
or interference? Psychological Review, 115, 544–576.
THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014
Downloaded by [the Bodleian Libraries of the University of Oxford] at 08:54 11 November 2014
AUTOMATIC SEMANTIC ENCODING
Oberauer, K., & Lewandowsky, S. (2013). Evidence
against decay in verbal working memory. Journal of
Experimental Psychology: General, 142, 380–411.
Paivio, A. (1986). Mental representations: A dual coding
approach. Oxford, England: Oxford University Press.
Poirier, M., & Saint-Aubin, J. (1995). Memory for
related and unrelated words: Further evidence on
the influence of semantic factors in immediate serial
recall. Quarterly Journal of Experimental Psychology,
48A, 384–404.
Quinn, J. G., & McConnell, J. (2006). The interval for
interference in conscious visual imagery. Memory,
14, 241–252.
Raye, C. L., Johnson, M. K., Mitchell, K. J., Greene, E.
J., & Johnson, M. R. (2007). Refreshing: A minimal
executive function. Cortex, 43, 135–145.
Romani, C., McAlpine, S., & Martin, R. C. (2008).
Concreteness effects in different tasks: Implications
for models of short-term memory. Quarterly Journal
of Experimental Psychology, 61, 292–323.
Rose, N. S., & Craik, F. I. M. (2012). A processing
approach to the working memory/long-term
memory distinction: Evidence form the levels-of-processing span task. Journal of Experimental Psychology:
Learning, Memory, and Cognition, 38, 1019–1029.
Rose, N. S., Myerson, J., Roediger, H. L., & Hale, S. (2010).
Similarities and differences between working memory
and long-term memory: Evidence from the levels-ofprocessing span task. Journal of Experimental Psychology:
Learning, Memory, and Cognition, 36, 471–483.
Schneider, W., Eschman, A., & Zuccolotto, A. (2002).
E-Prime user’s guide. Pittsburgh, PA: Psychology
Software Tools, Inc.
Sebastián, N., Martí, M. A., Carreiras, M., & Cuetos, F.
(2000). LEXESP. Léxico informatizado del español.
Barcelona: Ediciones de la Universidad de Barcelona.
Shivde, G. S., & Anderson, M. C. (2011). On the existence of semantic working memory: Evidence for
direct
semantic
maintenance.
Journal
of
Experimental Psychology: Learning, Memory and
Cognition, 37, 1342–1370.
Shivde, G., & Thompson-Schill, S. L. (2004).
Dissociating semantic and phonological maintenance
using fMRI. Cognitive, Affective, and Behavioral
Neuroscience, 4, 10–19.
Shulman, H. G. (1970). Encoding and retention of
semantic and phonemic information in short-term
memory. Journal of Verbal Learning and Verbal
Behavior, 9, 499–508.
Shulman, H. G. (1971). Similarity effects in short-term
memory. Psychological Bulletin, 75, 399–415.
Shulman, H. G. (1972). Semantic confusion errors in
short-term memory. Journal of Verbal Learning and
Verbal Behavior, 11, 221–227.
Towse, J. N., & Neil, D. (1998). Analyzing human
random generation behavior: A review of methods
used and a computer program for describing performance. Behavior Research Methods, Instruments and
Computers, 30, 583–591.
Vandierendonck, A., De Vooght, G., & Van der Goten,
K. (1998). Does random time interval generation
interfere with working-memory executive functions?.
European Journal of Cognitive Psychology, 10, 413–442.
Wagner, A. D., Pare-Blagoev, E. J., Clark, J., &
Poldrack, R. A. (2001). Recovering meaning: Left
prefrontal cortex guides controlled semantic retrieval.
Neuron, 31, 329–338.
Walker, I., & Hulme, C. (1999). Concrete words are
easier to recall than abstract: Evidence for a semantic
contribution to short-term serial recall. Journal of
Experimental Psychology: Learning, Memory, and
Cognition, 25, 1256–1271.
Whitney, C., Grossman, M., & Kircher, T. T. (2009).
The influence of multiple primes on bottom-up and
top-down regulation during meaning retrieval:
Evidence for 2 distinct neural networks. Cerebral
Cortex, 19, 2548–2560.
Whitney, C., Kirk, M., O’Sullivan, J., Lambon Ralph, M.
A., & Jefferies, E. (2011). The neural organization of
semantic control: TMS evidence for a distributed
network in left inferior frontal and posterior middle
temporal gyrus. Cerebral Cortex, 21, 1066–1075.
THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014
19
CAMPOY ET AL.
APPENDIX
Description of the word sets and comparison between concrete and abstract sets
Table A1. Word sets in Experiments 1 and 2
Downloaded by [the Bodleian Libraries of the University of Oxford] at 08:54 11 November 2014
Concrete words
Abstract words
Word property
Mean
SD
Range
Mean
SD
Range
t(24)
p
Concreteness
Imageability
Frequency
Familiarity
Phonemes
6.0
6.2
22.7
5.4
7.1
0.4
0.3
17.4
0.9
0.9
5.5–6.6
5.6–6.7
8.6–65.7
3.0–6.5
6–9
3.5
3.2
26.1
5.3
7.1
0.3
0.4
18.3
0.8
1.0
3.0–3.9
2.5–3.7
6.1–65.5
3.6–6.5
6–9
17.70
21.18
0.48
0.34
0.00
.00
.00
.63
.74
1.00
Note: Values of concreteness, imageability, word frequency, and familiarity were taken from the LEXESP database (Sebastián et al.,
2000). Concreteness, imageability, and familiarity ranged from 1 to 7. Word frequency is expressed in number of occurrences per
million words.
Table A2. Word sets in Experiment 3
Concrete words
Abstract words
Word property
Mean
SD
Range
Mean
SD
Range
t(138)
p
Concreteness
Imageability
Frequency
Familiarity
Phonemes
5.9
5.9
11.4
5.1
6.8
0.4
0.4
7.9
0.8
0.6
5.4–6.7
5.4–6.8
2.5–34.5
3.1–6.7
6–8
3.7
3.6
11.4
5.1
6.9
0.4
0.6
7.6
0.8
0.7
2.6–4.3
1.8–4.4
2.7–30.9
3.1–6.6
6–8
33.56
28.66
0.05
0.02
0.37
.00
.00
.96
.98
.71
Note: Values of concreteness, imageability, word frequency, and familiarity were taken from the LEXESP database (Sebastián et al.,
2000). Concreteness, imageability, and familiarity ranged from 1 to 7. Word frequency is expressed in number of occurrences per
million words.
20
THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2014
Download