Professional Documents
Culture Documents
A dissertation
W
submitted by
Doctor of Philosophy
in
PR
Psychology
TUFTS UNIVERSITY
May
2012
In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if material had to be removed,
a note will indicate the deletion.
W
IE
UMI 3512103
Copyright 2012 by ProQuest LLC.
EV
All rights reserved. This edition of the work is protected against
unauthorized copying under Title 17, United States Code.
PR
ProQuest LLC.
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, MI 48106 - 1346
Structure, Meaning, and Constituency in Visual Narrative Comprehension
W
IE
EV
PR
ii
Structure, Meaning, and Constituency in Visual Narrative Comprehension
Abstract
Narrative has been formally studied for at least two millennia, dating back
However, visual narratives in the form of sequential images have also been
W
printed in contemporary comic books and strips. Yet, compared with the study of
overall theory that sequential images at the narrative level are structured and
PR
processed analogously to sequences of words at the sentence level. The main idea
visual language used in comics, similar to the way that syntax organizes words
into coherent sentences. We focus here on two salient parts of this analogy. First,
this thesis will explore the idea that visual narrative comprehension involves a
iii
Structure, Meaning, and Constituency in Visual Narrative Comprehension
will briefly discuss the overall implications for the analogy between narrative
W
IE
EV
PR
iv
Structure, Meaning, and Constituency in Visual Narrative Comprehension
Acknowledgments
and Phil Holcomb for believing in a research project that had yet to find a place in
the broader studies of human cognition. My parents, Leigh and Lindsey Cohn,
years pursuing this work, as have my friends and colleagues, Laura Babbitt, Ian
Christian Blanche, Tali Ditman, Steve Maher, Ted M. Mason, Kathy Midgley,
Priya Mitra, Kana Okano, John Pacheco, Martin Paczynski, Agatha Trindade, Eva
W
Wittenberg, and Cherry Yum. Finally, my cousins Elizabeth Hall and Christian,
Henry and Emlen Lease are thanked for providing a place to set research aside, as
IE
are my students in the Tufts Soo Bahk Do Club: John Sotherland, Thomas Nikiper,
Martin Paczynski, Jason Rosenbaum, Negin Toosi, Madeline Gardner, Eric Sinski,
EV
Bender, and Adrienne Ing are thanked for helping with making stimuli and
running the experiment in Chapter 1. Suzi Grossman, Chelsey Ott, and Patrick
Bender are thanked for helping to make stimuli and running the experiments in
NICHD (HD25889) and NARSAD (with the Sidney Baer Trust), as well as
v
Structure, Meaning, and Constituency in Visual Narrative Comprehension
Table of Contents
Abstract iii
Acknowledgments v
Table of Contents vi
List of Tables ix
List of Figures ix
Preface 1
Chapter 1: Structure and Meaning in Visual Narrative Comprehension 4
Semantic approaches 4
Verbal narrative 4
Visual narrative 5
Structural Approaches 6
W
Verbal narrative 7
Interface of semantics and structure in verbal narrative 8
Narrative structure in sequential images
IE 9
Structure and semantics in processing 13
ERPs in language processing 14
ERPs in the visual-graphic domain 17
Processing semantics and structure in visual narrative 19
EV
Outstanding questions 21
Methods 27
Stimuli 27
Objective measures of coherence 30
PR
vi
Structure, Meaning, and Constituency in Visual Narrative Comprehension
W
Participants 85
Procedure 85
Data Analysis 86
Coherence ratings 86
Viewing times
IE 86
Results 90
Coherence ratings 90
EV
Viewing times 91
Disruptions vs. No-Disruptions 91
Effect of crossing a narrative constituent boundary 92
Effect of Disruption later in a sequence 93
PR
vii
Structure, Meaning, and Constituency in Visual Narrative Comprehension
W
IE
EV
PR
viii
Structure, Meaning, and Constituency in Visual Narrative Comprehension
List of Tables
List of Figures
W
Figure 1. Comic strip with prototypical narrative categories. 10
(Critical Panel+1). 40
ix
Structure, Meaning, and Constituency in Visual Narrative Comprehension
W
IE
EV
PR
x
Structure, Meaning, and Constituency in Visual Narrative Comprehension
W
IE
EV
PR
1
Structure, Meaning, and Constituency in Visual Narrative Comprehension
Preface
Narrative has been formally studied for at least two millennia, dating back
However, visual narratives in the form of sequential images have also been
printed in contemporary comic books and strips (Kunzle, 1973; McCloud, 1993).
W
Yet, compared with the study of discourse in verbal language, the study of
sequential image comprehension has been relatively impoverished. Just what are
IE
the structures motivating visual narratives and how are they processed?
overall theory that sequential images at the narrative level are structured and
processed analogously to sequences of words at the sentence level. The main idea
PR
visual language used in comics, similar to the way that syntax organizes words
into coherent sentences. We focus here on two salient parts of this analogy. First,
to the interaction between syntactic structure and semantics at the sentence level.
2
Structure, Meaning, and Constituency in Visual Narrative Comprehension
images convey far more information than words, perhaps even at the level of a
the analogy between visual narrative and sentences serves as a useful framework
to inform and interpret the study of sequential image comprehension, both for
W
their similarities and their differences.
First, this thesis will explore the idea that visual narrative comprehension
IE
involves a system of narrative structure and a system of semantic coherence that
implications for the analogy between narrative structure in sequential images and
syntax in sentences.
3
Structure, Meaning, and Constituency in Visual Narrative Comprehension
between structure and meaning. At the sentence level, theories have debated
whether sentences are guided more by syntax (e.g., Chomsky, 1957, 1965, 1995),
both (e.g., Culicover & Jackendoff, 2005; Jackendoff, 2002; Sadock, 1991).
Similar distinctions have been drawn beyond the level of individual sentences,
between narrative and meaning. While this distinction at the narrative level has
W
mostly been articulated for verbal discourse, we have argued that an analogous
relationship occurs between structure and meaning in the visual domain, most
IE
prominently in the sequential images used in comics (Cohn, In Press; Cohn,
Semantic approaches
Verbal narrative
4
Structure, Meaning, and Constituency in Visual Narrative Comprehension
(Bransford & Johnson, 1972; Brown & Yule, 1983; Schank & Abelson, 1977; van
Dijk, 1977) or through local connections between each sentence (Halliday &
Hasan, 1976; Hobbs, 1985; Kehler, 2002; Mann & Thompson, 1987). On the
dimensions of reference (Haviland & Clark, 1974), temporal and event structure
(Mandler, 1986; Speer & Zacks, 2005; Zwaan, 1996), causation (Black & Bower,
1979; Keenan, Baillet, & Brown, 1984; Trabasso, Secco, & van den Broek, 1984),
W
spatial structure (Clark, 1972; Linde & Labov, 1975; Morrow, Greenspan, &
Bower, 1987), and intentionality (Lichtenstein & Brewer, 1980; Schank &
IE
Abelson, 1977). While reading a discourse, readers integrate these domains
through a discourse, they continually update their situation model with new
PR
situation model with this new information, incurring a cost in processing (Zwaan,
Visual narrative
5
Structure, Meaning, and Constituency in Visual Narrative Comprehension
scenes (Cohn, 2003; McCloud, 1993; Saraceni, 2000, 2001; Stainbrook, 2003).
These transitions allow readers to inferentially “fill in the gaps” between adjacent
when an incoming panel introduces new information or does not share the
relations have not been directly tested in experimental settings on static visual
W
studies of film. While watching a film, viewers are consciously able to identify
changes in time, characters, and spatial location (Magliano, Miller, & Zwaan,
IE
2001; Zacks, Speer, & Reynolds, 2009), and are the most sensitive to changes
between full scenes—where film shots shift from one action to another at
EV
different times and spaces (Magliano & Zacks, 2011). Additionally, viewers can
between shots and by drawing on global scripts about film storytelling (Magliano,
Dijkstra, & Zwaan, 1996). Thus, like in verbal discourse, viewers of films draw
Structural Approaches
relationships between and across units. Some of the earliest theories of narrative
posited that narrative units play roles relating to a global structure. Dating back to
approaches agreed that narratives are organized into a basic canonical pattern
6
Structure, Meaning, and Constituency in Visual Narrative Comprehension
Verbal narrative
1957, 1965). Some approaches in this vein have modeled the structure of
conversational discourse (Clark, 1996; Hinds, 1976; Labov, 1972; Labov &
Waletzky, 1967). However, the generative theories of narrative that gained the
W
most attention were “story grammars” from the late 1970s. These models
evidence that participants do prefer stories organized into canonical orders. Much
of this research used memory paradigms that asked participants to recall aspects
PR
of stories that they just read. On the whole, participants were better able to recall
the order of events in a story when it was presented in canonical order than when
the sentences of the story were inverted (Mandler, 1978, 1984; Mandler &
DeForest, 1979), taken out of their temporal order (Mandler & Johnson, 1977), or
fully scrambled (Mandler, 1984). In fact, a correlation was found between the
accuracy of recalling a story and the degree to which it was altered away from the
canonical order (Stein & Nezworski, 1978). These studies supported that readers
prefer certain canonical narrative patterns that extend beyond the scope of general
7
Structure, Meaning, and Constituency in Visual Narrative Comprehension
Though story grammar research provided evidence that readers draw upon
story grammars set out to describe structural aspects of narratives, many critics
scale (Black & Wilensky, 1979; de Beaugrande, 1982). Indeed, the goal-directed
W
intentions and script-based events, just applied to stories (Bower, Black, & Turner,
1979; Lichtenstein & Brewer, 1980; Schank & Abelson, 1977). The use of scripts
IE
in particular could have been a confounding factor, because order of events in a
script could have been driving the effects of the narrative. For example, if a story
EV
semantic “restaurant script” (including the routine of dining and the elements
PR
involved) may have more of an impact on the recall of those events than the
information they extract from the narrative (Ohtsuka & Brewer, 1992).
(Brewer & Lichtenstein, 1981, 1982; Johnson-Laird, 1983; van Dijk, 1977), an
(Brewer, 1985; Brewer & Lichtenstein, 1981, 1982; Ohtsuka & Brewer, 1992)
emphasized that this separation allows for different mappings between narrative
8
Structure, Meaning, and Constituency in Visual Narrative Comprehension
and events. First, narratives can evoke different affective states in a reader based
actually show time travel, instead of merely presenting events out of order from
their actual instantiation. Brewer and colleagues found that participants are more
W
accurate at comprehending stories where the order of events is isomorphic to its
but from their interaction. Despite this insight, their research did not fully posit
just what the structure of narrative itself comprises outside of this interaction.
What would a formal model of narrative structure look like that also interfaces
with semantics?
Recently, Cohn (In Press) has proposed a theory of narrative structure for
sequential images that makes a clear division between narrative structure and
meaning. In this model, individual panels have categorical roles that are
9
Structure, Meaning, and Constituency in Visual Narrative Comprehension
this hierarchic structure will be addressed in depth in other studies (see Chapter 2),
here we will focus on narrative categories and the relationship between narrative
W
ES: [Event {Prep HOLD(BALL){Prep REACH}} {Head THROW} {Coda LOOK}]
which sets up the characters and interaction involved in the sequence. As in the
first panel, where the boy passively holds a bowling ball, these panels most often
depict passive states, since they function to simply introduce the characters. Next,
preparatory actions (such as the preparation to roll the bowling ball). The
narrative climaxes in the Peak, showing the main events of the sequence. In
Figure 1, the Peak is the penultimate panel, where the boy throws the ball. A
Release then shows the aftermath or coda of the event in the Peak, as in the final
panel of the sequence, where the ball stops rolling (to comedic effect).
10
Structure, Meaning, and Constituency in Visual Narrative Comprehension
this case, narrative categories map directly to events. Discourse theories have
proposed that prototypical mappings like this should be easier to comprehend than
narratives that are not isomorphic to the events they describe (Enkvist, 1981;
narrative are not always prototypically mapped, and it is important to stress that
W
semantic characteristics shown in a panel provide cues to its narrative category
only to be run over by the boys playing hockey. This sequence does follow the
climaxing in a Peak, and then ending with a Release. In this case, the semantics of
11
Structure, Meaning, and Constituency in Visual Narrative Comprehension
each panel do not match the prototypical correspondences for narrative categories
described above. At the start, the Establisher in this sequence does not show a
passive state. This panel depicts a puck in the process of sliding and Snoopy
surprised by it, but it still does functionally set up the interaction between them.
Next, the Initial is not a preparatory action—it shows Snoopy already engaged in
the process of chasing the puck. This is not necessarily an “initiating” event, but it
acts as a narrative Initial because of its relationship to the Peak. The subsequent
Peak panel does reach a climax of the narrative, but it does not show the
W
culmination of Snoopy’s actions that are prepared by the Initial. Rather, the boys
hiding, not, for example, Snoopy being run over on the ground as direct the
EV
consequence of the Peak’s actions. In this way, narrative categories do not always
map to prototypical events, and both content and context factor into the
PR
Thus, this approach allows for narrative structure and semantics to operate
between them. Given this theoretical framework, how might we find evidence for
narratives?
1
It is also worth noting that this sequence does not involve goal-directed actions. Is Snoopy’s goal
12
Structure, Meaning, and Constituency in Visual Narrative Comprehension
that the nature and constraints of narrative structure in sequential images and of
contain more information than words. Indeed, images are closer to conveying the
W
in looking for ways to differentiate narrative structure from semantic structure in
level.
EV
semantic structure at the sentence level extend back to the early days of
PR
psycholinguistics (e.g., Marks & Miller, 1964; Miller & Isard, 1963). Over the
because they directly index the underlying neural processes with excellent
temporal resolution. Several ERP effects have been correlated with semantic and
at the sentence level. We then describe how similar ERP effects appear to
13