Abstracts 2005 - The Psychonomic Society

Papers 279–285 Sunday Morning
sociations using conditional probability and global co-occurrence
similarity (from the HAL model). A priming experiment was conducted,
attempting to see what relationship (association strength, feature
similarity, or global co-occurrence similarity) explained the most
variance in RT. Results showed that after accounting for categorical
relatedness, only global co-occurrence similarity predicts additional
RT variance. The facts that these statistical relationships can predict
normative productions and explain RT differences that norms cannot
explain supports the argument that a representation based on these statistical
relationships forms the most comprehensive and parsimonious
theory of semantic representation.
8:40–8:55 (279)
Assessing Summaries With Human Judgments Procedure and Latent
Semantic Analysis. JOSÉ A. LEÓN, RICARDO OLMOS, & IN-
MACULADA ESCUDERO, Universidad Autónoma de Madrid, &
JOSÉ J. CAÑAS & LALO SALMERÓN, Universidad de Granada—
In this study, we tested a computer-based procedure for assessing very
concise summaries (50 words) using latent semantic analysis (LSA)
combined with four expert human judgments and two types of text
(narrative and expository). LSA was used to estimate semantic similarity
using six different methods: four holistic (summary–text, summary–
summaries, summary–expert summaries, and pregraded–ungraded summary)
and two componential methods (summary–sentence text and
summary–main sentence text). A total of 390 Spanish middle grade
students (14–16 years old) and six experts read a narrative or expository
text and later summarized it. Despite the fact that results supported
the viability of developing a computerized assessment tool
using human judgments and LSA, human reliability–LSA was higher
in the narrative text than in the expository, and LSA correlated more
with human content rates than with human coherence rates. Finally,
holistic methods were found to be more reliable than the componential
methods analyzed in this study.
9:00–9:15 (280)
Semantic Network Structure Appears in a Social Network in Dreams.
RICHARD J. SCHWEICKERT, Purdue University—In a semantic
network for words, each word is represented as a vertex, and two vertices
are connected if the corresponding words have similar meaning,
indicated by, say, appearing in the same category in Roget’s thesaurus.
Steyvers and Tenenbaum (2005) found that such networks have three
properties of a class of networks called small-world networks. Average
distance between two words is small, clustering is high, and the
number of connections for each vertex follows a power law. These
properties are of interest because they appear in structures such as the
World-Wide Web. Finding these properties in semantic networks suggests
that human memory has them as well. Evidence from sleep labs
indicates that dreams are largely a product of memory. If so, one
would expect to see small-world properties appearing in dreams.
Analysis is presented of the social network of characters appearing in
the dreams of an artist.
9:20–9:35 (281)
A Framework for Analyzing Sequence Learning. JOSEPH TZELGOV
& AMOTZ PERLMAN, Ben Gurion University of the Negev—We attempt
to characterize sequence learning in terms of automatic versus
nonautomatic processing and to apply this contrast independently to
knowledge acquisition and knowledge retrieval. In several experiments
on sequence learning, automaticity of acquisition and automaticity
of retrieval of the acquired knowledge were independently assessed.
We found that sequence learning order can be acquired under
all combinations of knowledge acquisition and knowledge retrieval.
In particular, at least in the case of the simple sequences we employed,
our finding applies when both the acquisition and the retrieval of
knowledge are strictly automatic—that is, when neither of them is part
of the task requirement—and are not beneficial to deliberate behavior.
The proposed framework has implications for the notion of sequence
learning and for the investigation of learning in general.
44
Attention III
Dominion Ballroom, Sunday Morning, 8:00–10:00
Chaired by Christian N. L. Olivers, Vrije Universiteit Amsterdam
8:00–8:15 (282)
Spreading the Sparing: Against a Limited-Capacity Account of the
Attentional Blink. CHRISTIAN N. L. OLIVERS, & STEFAN VAN DER
STIGCHEL, Vrije Universiteit Amsterdam, & JOHAN HULLEMAN,
Hull University—The second of two targets embedded in a stream of
distractors is often missed when presented in close temporal proximity
to the first—a phenomenon known as the attentional blink. It is thought
that the first target consumes limited resources that are then denied to
the second. Against this limited-capacity account, Di Lollo, Kawahara,
Ghorashi, and Enns (2005; Psychological Research, 69, 191–200)
found that three targets can be accurately detected as long as a distractor
does not intervene. Extending this work, we demonstrate that
(1) presenting even more targets does reveal capacity limitations, but
these limitations are additive with the attentional blink effect; and
(2) targets can be spared from the blink at any point in time whenever
they are preceded by another target. The results indicate that temporal
attention is much more dynamically controlled than has been previously
thought.
8:20–8:35 (283)
The AB Is Due to an Exogenous Disruption of Endogenous Control,
Not to Resource Depletion. VINCENT DI LOLLO, Simon Fraser
University—Identification of the second of two targets is impaired if
it is presented less than about 500 msec after the first. Theoretical accounts
of this second-target deficit, known as attentional blink (AB),
have relied on some form of limited attentional resource that is allocated
to the leading target at the expense of the trailing target. This
resource depletion account is disconfirmed by the outcome of experiments
in which three targets were used instead of the conventional
two. These experiments indicate that the AB arises not from resource
depletion but from a disruption of endogenous control over the current
attentional set. The disruption is triggered exogenously when a distractor
is presented while the central executive is busy orchestrating
the processing of the first target. This leads to impaired identification
of any trailing targets presented within about 500 msec—namely, before
endogenous control can be reestablished.
8:40–8:55 (284)
Space-Based and Object-Based Visual Attention in Overlapping
Tasks. HAGIT MAGEN, Princeton University, & ASHER COHEN,
Hebrew University of Jerusalem (read by Asher Cohen)—When subjects
perform two tasks (T1 and T2) in close temporal succession, performance
of T2 is typically severely limited. However, much evidence
suggests that some aspects of T2 can be performed concurrently with
T1. In our previous research with a visual T2, we showed that this performance
limitation can be explained to a large extent by assuming a
modular architecture of the visual system. In this talk, we examine
whether visual attention can operate on T2. A previous paper showed
that spatial attention can be used for T2 concurrently with the performance
of T1. We present evidence, using the classical Duncan task,
that object-based attention can also be used for T2. We describe a
model that assumes different roles for space- and object-based attention
systems during overlapping tasks and present an additional experiment
with the PRP paradigm that confirms the prediction of this
model.
9:00–9:15 (285)
Sparing and Blinking in the Simultaneous-Type/Serial-Token Model
of the Attentional Blink. BRAD WYBLE, PATRICK CRASTON, &
HOWARD BOWMAN, University of Kent, Canterbury—A previously
published computational model of the attentional blink (Wyble & Bowman,
2004) has given rise to predictions about the time course of the
U-shaped T2 performance curve that is characteristic of this phenom-