THE ATTENTIONAL DISENGAGEMENT MODEL OF THE MISSING-LETTER EFFECT: A TEST OF THE ATTENTIONAL BEAM
by
AMANDA LALANDE
Thesis submitted in partial fulfillment of the requirement
for the degree of Master of Arts in Psychology (M.A.)
School of Graduate Studies
Laurentian University
Sudbury, Ontario
© AMANDA LALANDE, 2011 Library and Archives Bibliotheque et 1*1 Canada Archives Canada Published Heritage Direction du Branch Patrimoine de I'edition
395 Wellington Street 395, rue Wellington OttawaONK1A0N4 Ottawa ON K1A0N4 Canada Canada
Your file Votre reference ISBN: 978-0-494-82018-6 Our file Notre reference ISBN: 978-0-494-82018-6
NOTICE: AVIS:
The author has granted a non L'auteur a accorde une licence non exclusive exclusive license allowing Library and permettant a la Bibliotheque et Archives Archives Canada to reproduce, Canada de reproduire, publier, archiver, publish, archive, preserve, conserve, sauvegarder, conserver, transmettre au public communicate to the public by par telecommunication ou par I'lnternet, preter, telecommunication or on the Internet, distribuer et vendre des theses partout dans le loan, distribute and sell theses monde, a des fins commerciales ou autres, sur worldwide, for commercial or non support microforme, papier, electronique et/ou commercial purposes, in microform, autres formats. paper, electronic and/or any other formats.
The author retains copyright L'auteur conserve la propriete du droit d'auteur ownership and moral rights in this et des droits moraux qui protege cette these. Ni thesis. Neither the thesis nor la these ni des extraits substantiels de celle-ci substantial extracts from it may be ne doivent etre imprimes ou autrement printed or otherwise reproduced reproduits sans son autorisation. without the author's permission.
In compliance with the Canadian Conformement a la loi canadienne sur la Privacy Act some supporting forms protection de la vie privee, quelques may have been removed from this formulaires secondaires ont ete enleves de thesis. cette these.
While these forms may be included Bien que ces formulaires aient inclus dans in the document page count, their la pagination, il n'y aura aucun contenu removal does not represent any loss manquant. of content from the thesis.
1+1 Canada
AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Abstract
When reading, participants miss more target letters in frequent function words compared
to rare content words. The Attentional Disengagement model proposes the missing-letter
effect is due to the precise timing of disengagement of the attentional beam. This study
constituted the very first systematic exam of the attentional beam within the context of
the missing-letter effect. More precisely, this study sought to examine the scope of
processing and properties of the attentional beam. Division signs positioned along the top
and bottom of lines of text were transformed into minus signs within the beam. This was
expected to yield a differential rate of omission as a function of probe delay and word
type. The results revealed a reverse missing-letter effect; a higher rate of omission for
probes associated with content versus function words. This finding is the first of its kind
within the entirety of the missing-letter effect field. We propose the scope of processing
of the attentional beam varies as a function of the cognitive load associated with a word.
On the one hand, function words represent a light cognitive load which permits the
expansion of the beam and facilitates probe detection. On the other hand, content words
represent a heavy cognitive load which constrains the expansion of the beam and impairs probe detection.
in AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Acknowledgements
This thesis would not have been possible without the help, support, and
contributions of many. I would like to acknowledge the following individuals for the
important roles they played in the completion of this research and written document.
Firstly, I would like to thank my Master's thesis supervisor Dr. Annie Roy-
Charland. Thank you for your constant guidance and support throughout the entirety of
this project. I am also grateful for your enthusiasm, your encouragement, your interest in
my professional development, and even your early morning phone calls! Most of all,
thank you for the endless hours of revisions, for enduring my preoccupation with details,
and for allowing my 'wordiness' these past two years, as we so often put it! It all paid
off!
I would also like to thank my committee members Dr. Joel Dickinson and Dr.
Melanie Perron. Dr. Dickinson, thank you for your unique perspective and contributions to this project, and for allowing me to present my preliminary research findings in your class. Moreover, thank you for always ensuring that I remained on the road to success and minimized all unnecessary distractions throughout graduate school! Dr. Perron, thank you for your feedback and constructive criticism. You always delivered in with a friendly and approachable smile. To the both of you, I am grateful for your support and for all of your suggestions and revisions.
Thank you to my parents for your constant support and for the work ethic that you have instilled in me. I credit much of my academic success to you. Also, thank you for
iv AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM allowing me to crowd your otherwise empty nest these past two years. I promise I won't move in again!
A special thank you to my sister Joelle, to my dear friend Katherine, and to my boyfriend Angelo. Your unwavering support and understanding throughout the entirety of my academic career has meant so much to me. Thank you for always being my biggest fans!
I also want to thank Beth Emptage and the rest of my graduate colleagues for all of their support these past two years. Another important thank you to Jessica Boulard for her interest and special contribution to this project. Finally, I would like to thank all faculty members, graduate students, undergraduate students, and research assistants who are members of the Cognitive Health Science Research Laboratory. You will all be missed!
v AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Table of Contents
Abstract iii
Acknowledgements iv
Table of Contents vi
List of Tables viii
List of Figures ix
The Attentional Disengagement Model of the Missing-Letter Effect: A Test of the
Attentional Beam 1
History of the Missing-Letter Effect 2
Unitization Account/Processing Time Hypothesis 2
Structural Account 5
Guidance-Organization (GO) Model 7
Attentional Disengagement (AD) Model 15
The Attentional Beam 19
The Attentional Beam in Reading 25
Present Study 27
Hypotheses 33
Methodology 36
Participants 36
Material 36
Apparatus 37
Procedure 38
vi AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Results 40
Probability of Fixation 41
Omissions 44
Response Latencies 50
Fixation Durations 56
Word Frequency 63
Discussion 63
Overview 63
Reverse Missing-Letter Effect 66
Probe Delay Conditions 73
Limitations 75
Conclusion/Future Direction 79
References 81
vn AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
List of Tables
Name Page
Table 1 .Means and Standard Deviations Probabilities of Fixation of the Target Region of the Des and Pour/cour Texts as a Function of Probe Delay and Word Role 42
Table 2. Means and Standard Deviations for Omissions of the Target Region of the Dem and Pour/cour Texts as a Function of Probe Delay and Word Role 45
Table 3. Means and Standard Deviations for Omissions of the Target Region of the Des
Text as a Function of Fixation Status, Probe Delay, and Word Role 48
Table 4. Means and Standard Deviations for Response Latencies of the Des and
Pour/cour Texts as a Function of Probe Delay and Word Role 51
Table 5. Means and Standard Deviations for Response Latencies of the Des Text as a
Function of Fixation Status, Probe Delay, and Word Role 54
Table 6. Means and Standard Deviations for Fixation Duration Measures of the Des Text as a Function of Detection Status, Probe Delay, and Word Role 58
Table 7. Means and Standard Deviations for Fixation Duration Measures of the Des and
Pour/cour Texts as a Function of Detection Status and Probe Delay 61
vni AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
List of Figures
Name Page
Figure 1. Example of the manner in which text is unitized and processed in parallel at
several different levels of analysis (Healy, 1994) 3
Figure 2. Example of the manner in which whole word-processing time distribution is
split into two component distributions; omissions and detections (Roy-Charland et al.,
2009) 14
Figure 3. Example of the two superimposed letter streams used in Muller and Hubner's
(2002) study involving the attentional spotlight 23
Figure 4. Example of method to be used in Experiment 1 29
Figure 5. Probabilities of fixation in the Des text 43
Figure 6. Probabilities of fixation in the Pour/cour text 43
Figure 7. Proportion of omissions in the Des text 46
Figure 8. Proportion of omissions in the Pour/cour text 46
Figure 9. Proportion of omissions in the Des text as a function of fixation status 49
Figure 10. Response latencies in the Des text 52
Figure 11. Response latencies in the Pour/cour text 52
Figure 12. Response latencies in the Des text as a function of fixation status 55
Figure 13. Fixation duration measures in the Des text 62
Figure 14. Fixation duration measures in the Pour/cour text 62
ix AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
The Attentional Disengagement Model of the Missing-Letter Effect: A Test of the
Attentional Beam
Reading is a complex cognitive human process resulting from the coordination of
different functions from different brain regions (Huey, 1908; Rayner & Pollatsek, 1989;
Reichle, Warren, & McConnell, 2009). Many recent studies have sought to define the
precise cognitive mechanisms underpinning reading (e.g. Corcoran, 1966; Saint-Aubin &
Poirier, 1997; Rayner, 1998; Healy, 1976; 1994; Moravcsik & Healy, 1995; Greenberg &
Koriat, 1991; Koriat & Greenberg, 1994; Saint-Aubin, Roy-Charland, & Klein,
2003;Greenberg, Healy, Koriat, & Kreiner, 2004; Roy-Charland & Saint-Aubin, 2006;
Roy-Charland, Saint-Aubin, Klein, & Lawrence, 2007; Roy-Charland, Saint-Aubin,
Lawrence, & Klein, 2009). One particularly interesting phenomenon that has emerged
from reading research is known as the missing-letter effect (MLE). Using a simple letter-
cancellation task, Corcoran (1966) was the first to discover, among other findings, that when participants were asked to read a prose passage and manually cross out a target
letter, they made more omissions on the target letter when it was embedded in a high-
frequency function word such as "the", compared to when it appeared in a low-frequency content word.
Many studies have since provided converging evidence for Corcoran's (1966) unexpected discovery of the MLE, and various models have been proposed to explain this curious, yet well-replicated phenomenon. The most prominent of these models, which will be reviewed in detail below, include the Unitization Account/Processing Time
Hypothesis (Healy, 1976; 1994; Moravcsik & Healy, 1995), the Structural Account
1 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
(Greenberg & Koriat, 1991; Koriat & Greenberg, 1994), and the Guidance-Organization
(GO) Model (Greenberg et al., 2004). The central aspects of each model have been the
focus of much testing, research, and debate, and while each holds some merit in
explaining the MLE, none thus far have been completely successful in accounting for the
MLE across all conditions. In this vein, researchers have proposed the most recent
addition to the missing-letter effect field; the Attentional-Disengagement (AD) Model
(Roy-Charland et al., 2007; 2009). The main goal of the present research will be to test
the pre-lexical attentional mechanism at the core of the AD model. More precisely, the
present thesis will explore the concept of the attentional beam (Rayner, 1998).
History of the Missing-Letter Effect
Unitization Account/Processing Time Hypothesis. Using a simple letter-
detection paradigm, Healy (1976; 1994; Moravcsik & Healy, 1995) was the first to
explicitly test a series of explanations for the MLE, and then later elaborate what is now
widely referred to as the Unitization Account/Processing Time Hypothesis. This model
proposes that, during normal reading, text is unitized and processed in parallel at several
different levels of analysis (Healy, 1994). These include the visual features of a word,
and the letter, syllable, word, or supraword levels (Figure 1). One crucial factor,
according to this model, is that once a higher-level reading unit such as a word has been
identified by the reader, there is an interruption and cessation of lower level operations,
such as the processing of that word's constituent letters. Importantly, this interruption occurs regardless of whether or not lower-level processing is complete. Another central assumption of this model is that high-frequency words are unitized and processed more
2 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Supro- word [of rhej j— level
'*%\ r' %. Wort r e levei C , JjJ2]Jl) 8iM
Sylloble level "S3
Letter v level 0 0 H
Feature II 1 level
AcHiol of the reodi ng unit I n py I
Figure 1. Example of the manner in which text is unitized and processed in parallel at several different levels of analysis (Healy, 1994).
3 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
rapidly than low-frequency words. Specifically, the high level of familiarity associated
with a frequent word facilitates the execution of higher-level processes, such as the
identification of that word. Consequently, the processing of a target letter embedded in a
frequently-occurring word is interrupted more often than when the target letter appears in
a low-frequency word. The MLE occurs because high-frequency words rapidly drive
identification at the word level, which then results in the interruption of lower-level
mechanisms such as the processing of a word's constituent letters. This interference
reduces the time available for letter processing and detection, thereby resulting in the
reader's "missing" of the target letter.
Whereas Corcoran (1966) only hypothesized possible explanations in determining
the MLE, many of Healy's (1976; 1994; Moravcsik & Healy, 1995) experimental designs
enabled her to isolate, manipulate, and explicitly test word frequency as the critical factor
in her Unitization Hypothesis. One such example includes an experiment conducted by
Moravcsik and Healy (1995). Participants were presented with identical versions of a passage except for the critical test word which was the common pronoun "they" in one version, and the less common "thou" in another. The task required they read and circle the target letter "h" each time it appeared in a word. Not surprisingly, participants made significantly more letter-detection errors in the version containing the more frequent
"they" pronoun, versus the one containing the "thou" pronoun. This finding is best explained in terms of word frequency since both critical test words "they" and "thou" were matched for word meaning, word length, target letter location within the target- containing word, and syntactic roles within the text. In other words, word frequency was
4 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
the primary factor causing the MLE since it was the only dimension upon which the
critical test words differed. This experiment is one of many within the Unitization
Account/Processing Time Hypothesis which has provided direct evidence for the effect
of word frequency on the occurrence of the MLE.
Structural Account. The Unitization Account/Processing Time Hypothesis
(1976; 1994; Moravcsik & Healy, 1995) is certainly noteworthy in that it offers a
straightforward interpretation of the role of word frequency as it relates to the MLE.
However, one drawback to this early model is that it does not address the importance of
the grammatical role of words, another critical factor in determining the MLE. In this
context, Koriat and Greenberg (1994) have proposed yet another model; the Structural
Account. This model proposes that, during normal reading, sentence structure is first
extracted by the reader. This is done by identifying function words which serve as
structural anchors within the sentence. Once a structural frame or skeleton has been
established by these anchors, function words recede to a sort of cognitive background,
and sentence meaning is then derived by the reader. The processing of meaning is
achieved only after function words recede in this figure-ground manner and the reader's
attention is shifted to semantically rich content words within the sentence.
According to the Structural Account, the MLE is not due to the simple fact that
frequent function words are unitized and processed more rapidly than rare content words.
Instead, this model proposes that omissions on a target letter are due to the different syntactic or linguistic roles of function and content words. Highly familiar or frequent function words serve as structural anchors within a sentence. These quickly recede to the
5 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM background after a sentential frame or, in simpler terms, the framework of the sentence is established, in an effort to allow the reader to then derive meaning. Therefore, if a target letter is located within a function word (or structural anchor), it quickly becomes less accessible for letter processing and detection because it is dismissed to pave the way for meaning. This then results in high rates of omission. The detection of target letters embedded in content words, on the other hand, is more easily accomplished due to the fact that semantic analysis is the final step in text processing. In these cases, the reader is able to accurately detect the target letter for the content word that remains in the cognitive forefront.
Several lines of research support the idea of a grammatical influence on the MLE
(Greenberg & Koriat, 1991; Koriat & Greenberg, 1994; Saint-Aubin & Poirier, 1997).
One example of this includes a clever experiment conducted by Saint-Aubin and Poirier
(1997). Participants were asked to read a passage and circle all instances of the letter "r".
This passage contained the French word "or", however it was used both in a structure- supporting conjunctional capacity, and as a noun. The fact that the word "or" has approximately the same frequency in the French language in both grammatical capacities allowed the researchers to isolate and test the syntactic role associated with each instance of the word. The idea was that the conjunctional "or" acted as a structural anchor in the sentence, whereas the noun "or" did not carry any structural information. Interestingly, the results of this study revealed that participants made significantly more omissions on the target letter in the word "or" when it was used in a conjunctional sense, versus when it was used as a noun, despite the fact that both their usages have similar frequencies.
6 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
According to the Structural Account and as discussed in the above, this is due to the fact that there is a cognitive fading of conjunctions, or structural anchors, in an effort to allow the transition from structure to meaning.
Guidance-Organization (GO) Model. According to both the Unitization
Account/Processing Time Hypothesis and the Structural Account, participants make a disproportionate amount of letter detection errors in frequent function words. The former model postulates that this is due to the effect of word frequency. In contrast, the latter model proposes that this is due to the effect of the grammatical role of function words.
The abundance of empirical evidence supporting each view suggests that neither model on its own can account for MLE across all conditions. As such, researchers have proposed an additional model which seeks to link the central assumptions of each of these rival accounts; the Guidance-Organization (GO) Model (Greenberg, Healy, Koriat, &
Kreiner, 2004).
According to the GO model, during normal reading, one of the very first processing steps involves the reader's attempt at establishing structure within a sentence.
As in the Structural Account, this typically involves the identification of "functors"; highly familiar, frequent function words which serve as structural anchors. Once a tentative structural frame is constructed by means of these "functors", attention is shifted to content words from which the reader then acquires meaning about the sentence. Unlike the Structural Account which does not elaborate any specific mechanism regarding the identification of "functors", the GO model proposes that it is the unitization of frequent function words which acts as the critical factor in this process. More specifically,
7 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM function words are unitized and processed more rapidly than content words which results in their usage as "functors" or structural anchors within the sentence.
Two main elements or assumptions are central to the GO model; 1) the integrated effect of word frequency and role and 2) processing time as the proximal cause and word frequency and role as the distal causes of the MLE. In light of the fact the GO model is an assimilation of the two competing views described above, it is not surprising that it also predicts an interaction between the central elements associated with each; word frequency and word function. In other words, there is an integrated effect of word frequency and role. More specifically, the GO model predicts higher rates of omission for function words than for content words, but only in connection with high-frequency words. This is due to the fact that, as previously discussed, the GO model credits word frequency as the mechanism responsible for the identification of "functors" or structural anchors. Thus, only frequent function and content words should differ in terms of omission rates. Rare function and content words should yield the same approximate rates of omission for, in these cases, frequency is not acting upon word function. These predictions were empirically tested and supported previous research by Roy-Charland and Saint-Aubin (2006). In their study, participants were asked to read a French-language passage and to detect the target letter "s" each time it appeared in a word. Target- containing words within this study included the high-frequency function word "depuis"
(since), the high-frequency content word "cours" (course), the low-frequency function word "hormis" (apart from), and the low-frequency content word "intrus" (intruder). As predicted, their results revealed a significant difference between rates of omission related
8 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM to the frequent function word "depuis" (since) and the frequent content word "cours"
(course). In contrast, they reported no significant difference in terms of rates of omission between the rare function word "hormis" (apart from) and the rare content word "intrus"
(intruder).
As stated above, the GO model proposes that the proximal cause or main mechanism underlying the MLE is the time available for processing a word. This aspect of the GO model is derived directly from the Unitization Account/Processing Time
Hypothesis, upon which the GO model was constructed. According to the former, frequent words are more rapidly identified, resulting in the interruption and cessation of lower-level mechanisms, such as letter processing. This then leads to the reader's
"missing" of the target letter. In line with this view, the GO model assumes that omissions occur when an insufficient amount of time is allocated to the processing of a word, thereby reducing the time available for letter processing.
In addition, the GO model further stipulates that word frequency and word function are distal causes underlying the MLE, for each influences the speed at which a word is processed. On the one hand, frequent words possess a highly familiar visual configuration which drives identification and, thus, processing at the word level. On the other hand, frequent function words, which are typically used as "functors" in establishing sentence structure, are always identified and processed more rapidly than neighbouring content words. As stated, according to the GO model, the proximal cause or main mechanism underlying the MLE is the time available for word processing. In this architecture, processing is accelerated for frequent words, and especially frequent
9 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM function words, for the properties associated with both frequency and function in this direction drive the speed of processing.
The GO model is certainly persuasive in that it integrates the fundamental aspects of both the Unitization Account/Processing Time Hypothesis and the Structural Account.
In addition, empirical research has supported some of its major predictions (Roy-
Charland & Saint-Aubin, 2006). However, a series of recent experiments have discredited what the GO model advances as the main mechanism underlying the MLE; the time available for processing a word (Roy-Charland et al., 2007; 2009). That is to say, findings have emerged from these studies which cannot be explained in terms of the Processing
Time Hypothesis, the center stage of the GO model. The researchers of these studies have devised two separate, yet complimentary techniques to empirically test the Processing
Time Hypothesis: 1) measuring fixation durations and 2) comparing the standard deviation of these fixation durations. The results of each method have provided evidence contrary to the Processing Time Hypothesis.
Fixation duration measures included first-fixation duration, single-fixation duration, gaze duration, and total fixation time. First-fixation duration consists of the time spent on a word, regardless if this time represents a single fixation or the first in a series of subsequent fixations. Single-fixation duration consists of the time spent on a word when it is only fixated once. Gaze duration represents the sum of all fixations on a word that occur before a saccade to another word takes place. Importantly, gaze duration is dependent on the reader entering the word from left to right. Finally, total fixation time
10 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM represents the sum of all fixations, including regressions and forward fixations on the word. (Rayner, 1998; Roy-Charland et al., 2007).
According to GO, if processing time is the main element underlying the MLE, then fixation durations for words should be longer for detections versus omissions for the latter represent faster processing times. This is on account of the fact that omissions on a target letter occur when there is an interruption (truncation) in letter processing due to activation and more rapid processing at the word level. Likewise, detections on a target letter typically take place when identification at the word level is longer. However, one alternative possibility or confounding factor investigated by Roy-Charland and colleagues (2009) is that the additional time spent on a word in cases of detection is not on account of added processing, but is the result of the planning and subsequent production of a motor (detection) response.
In their study, participants were divided into two separate groups and presented with a passage. The task required that the first group of participants read the passage at their normal reading speed and click on the mouse button each time they observed the letter "d" in a word. The second group of participants was presented with the same passage. However, their task required that they read at their normal reading speed and press the mouse button each time they observed the letter "p" in a word. The first group represented the experimental condition for fixation durations associated with omissions and detections were based on their results. The second group, however, was only implemented in an effort to measure the overall fixation durations on the critical "d" words within the passage. In other words, the second group represented the baseline
11 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM condition. Since the letter "p" never appeared in any of the critical "d" words, the fixation durations of this baseline group provided a benchmark against which the researchers could then compare the fixation durations for detections and omissions of the experimental group.
In accordance with processing time, the results revealed longer fixation durations for detections versus omissions. That is, participants spent more time on a word when they detected the target letter, compared to when it was missed. Based on this finding which supports processing time, longer fixation durations for detections versus the baseline condition would be expected, as well as shorter fixation durations for omissions versus the baseline condition. This is due to the fact that the GO model predicts the longest fixation durations when the reader detects the target letter (detections), followed by when the reader processes the entire word without the occurrence of a detection or omission (baseline condition), and, finally, by when the reader omits the target letter
(omissions). As predicted, fixation durations were longer for detections versus the baseline condition. But, the results also revealed that, contrary to the above, fixation durations were never significantly shorter for omissions compared to the baseline condition. In other words, participants spent the same approximate time fixating a word when they "missed" the target letter (omissions) and when the entire word was processed without an omission or detection (baseline condition).
This finding does not seem to fit within the GO model for, within the context of processing time, the latter states that, due to truncation in letter processing, the time associated with omissions should be significantly less than that associated with the
12 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM processing of the entire word distribution. However, as stated above, this experiment yielded a somewhat opposing pattern of results. As such, the processing time hypothesis associated with detections must be rejected due to the unexpected findings involving the processing times of omissions and the baseline condition. In addition, the alternative possibility that fixation durations are comprised of both the processing of the word and the planning and execution of a motor (detection) response seems more parsimonious in explaining fixation durations for detection responses. According to the researchers of this study, these findings certainly do not fit within the framework of processing time in determining the MLE. However, some difficulty still persisted at the time for it did not seem reasonable to reject an entire model based on a null and somewhat indirect result.
In an effort to then further refute the mechanism of processing time within the GO model, Roy-Charland and colleagues (2009) examined the standard deviation of fixation duration measures across all conditions. According to the GO model, fixation durations for detections and omissions each represent a subset of the overall processing time required to identify a word (Figure 2). Consequently, their respective standard deviations should not exceed the standard deviation for the overall processing of a word. In other words, the combination of the two subsets should yield the same approximate standard deviation as the overall processing. Importantly, the standard deviation regarding the overall processing was measured using the baseline condition described above. The results reported by Roy-Charland and colleagues (2009) were once again inconsistent with processing time within the GO model. The standard deviation of fixation durations for omissions and detections was never significantly smaller than that obtained in the
13 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
-*• Letter-Processing Time
*• Time
Omissions Detections
Figure 2. Example of the manner in which whole word-processing time distribution is split into two component distributions; omissions and detections (Roy-Charland et al., 2009).
14 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM baseline condition. As a matter of fact, the standard deviation of all fixation duration measures for detections was significantly larger than that reported in the baseline condition. In addition, the standard deviation of total fixation time was significantly larger for omissions versus the baseline condition.
Processing time as stipulated by the GO model states that longer fixation durations cause detections for there is no truncation; in other words, there is enough time available for participants to process a word in its entirety, including its constituent letters.
However, the fact that there is no significant difference in fixation duration between omissions and the overall processing of a word, as predicted by truncation models, has led researchers to conclude that this idea of processing time must be rejected. Most importantly, fixation times for detections and omissions are more variable than those associated with the overall processing. Again, this finding appears to be inconsistent with predictions from truncation models. Instead, it seems to fit well within the framework of the alternative possibility described above; that detections cause longer fixation durations for these involve the planning and execution of a motor (detection) response. Considering that processing time represents such an integral part of not only the theoretical integration offered by GO, but of its constituent models, these results essentially invalidate all previous models as well.
Attentional Disengagement (AD) Model.
The lack of a working model in explaining the MLE then compelled researchers to propose yet another alternative; the AD model. The AD model is unique within the realm of the missing-letter effect due to its quantitative nature. Importantly, it has
15 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM provided researchers with an original and empirical foundation upon which they have been able to systematically test previous models and deduce that the AD model alone can account all the findings within MLE literature, whereas previous models cannot.
According to the central tenets of this model, all instances of the MLE can be understood in terms of the manner in which attention is allocated during normal reading.
Specifically, the AD model proposes that reading results in an attentional beam being engaged on a word or a target-containing lexical unit. As time passes, information accumulates regarding the word or unit, including the possible presence of a target letter.
This information then helps to create an internal mental representation of the word. Once the processing of the word is complete, the attentional beam is shifted in a serial fashion to the spatial location of the next word or unit. The disengagement of attention is such that the information gathered to produce the word's internal mental representation begins to deteriorate. This decrease in information then results in the overall degradation of the word's mental representation.
In this context, the AD model proposes that the MLE is due to the precise timing of attentional disengagement. Attention is engaged on a word during normal reading until the word is identified and comprehended by the reader. As discussed previously, attention moves in a serial fashion to the next word in the text only once this process is complete.
However, the temporal extent to which attention is engaged on a word or target- containing unit varies as a function of word role and frequency. In cases where attention is prematurely disengaged, there is sometimes an insufficient amount of time available to gather information beyond the overall aspects of the word, such as its meaning. More
16 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM precisely, there is limited time to access deeper and more complex knowledge of the word, such as information pertaining to the word's constituent letters. This then results in a shortage or lack of feature-containing information which produces a mental representation of the word of a lesser quality which, in addition, more rapidly deteriorates at the timing of disengagement. This, in turn, results in greater difficulty in identifying a possible target letter and, hence, higher rates of omission on that target letter.
One important feature of the AD model is that it does not deny the role of lexical features such as word frequency or word function in accounting for the MLE; instead, it proposes that these features influence the temporal extent to which attention is allocated.
That is, attention remains engaged for longer periods of time on some words, and shorter periods on others, due to word frequency, word role, or a combination thereof. On the one hand, frequent words are identified and processed more rapidly than rare words due to their high familiarity. Attention is therefore disengaged more rapidly from frequent words, resulting in higher rates of omission. On the other hand, readers more rapidly identify and process function words over content words, on the basis that they often serve as structural anchors within a sentence and are also easier to predict within a sentence. As such, attention is quickly disengaged from function words, also resulting in higher rates of omission.
Finally, the AD model differs from its predecessors in one other noteworthy manner. All previous models assume that omissions on a target letter occur due to the interruption (truncation) of letter processing. This idea is particularly well illustrated in
Healy's Unitization Account/Processing Time Hypothesis (Healy, 1976; 1994;
17 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Moravcsik & Healy, 1995). According to truncation models, once the word or target- containing unit has been identified by the reader, letter processing can no longer persist and detections can therefore no longer occur. Simply put, word identification signals the complete cessation of letter processing, altogether eliminating the possibility of detection on the target letter. In comparison, according to the AD model, there is no clear and concise interruption of letter processing; instead, it appears that attentional disengagement serves to signal the beginning of a gradual decay in perceptual information relating to a word. Consequently, detections are thought to still be able to occur on the heels of attentional disengagement for, at this time, not enough perceptual decay has taken place to inhibit or impair letter detection accuracy. That is to say, detection accuracy depends on the extent to which perceptual decay has taken place and, importantly, on the amount of time that attention was engaged. Thus, the crucial element of this model is the timing of attentional disengagement.
As of yet, this element has not been systematically tested for the AD model is still fairly new. However, one recent study by Saint-Aubin, Klein, Deacon, and Thompson
(2010) has provided support for the attentional processes of the AD model with respect to lexical features, such as colour. In their study, participants were asked to detect a specific colour in a text which was comprised of letters of various colours. Interestingly, a missing-colour effect was reported. These findings provide empirical support for the concept of an attentional beam within the context of the MLE. What is more, they suggest that processing is homogenous for both lexical and non-lexical features within the beam. What remains to be tested within this framework, however, is how the
18 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM mechanisms of this model function with respect to non-lexical and reading irrelevant features, such as boundaries and symbols. Furthermore, it will be of interest to investigate the scope of processing of the beam when irrelevant features are not directly letter or lexically-bound.
The Attentional Beam.
Attention refers to the mental act of allocating cognitive resources to something internal or within our environment, at the expense of simultaneous and competing stimuli
(James, 1890). Posner (1980) and Posner and Cohen's (1984) early work involving the covert orienting of attention essentially gave rise to the now popular and widely investigated idea of the attentional beam. Briefly, the attentional beam is thought to act by enhancing the processing of elements which are illuminated and contained within its surround. In their series of experiments, Posner (1980) and Posner and Cohen (1984) asked participants to maintain visual fixation on a central element (e.g. fixation cross) of a display screen. An exogenous cue (e.g. a flash) then appeared somewhere on the screen, indicating a specific direction or region. The cue was then followed by a target stimulus.
The reaction times of participants in detecting the target were significantly faster when it appeared in the cued versus the uncued region. The researchers attributed this to the fact that the exogenous cue had engaged the internal "mental spotlight" or attentional beam of the participants at the cued location before the onset of the target. Consequently, processing was enhanced, resulting in a benefit, once the target appeared at this location.
In comparison, processing was delayed, resulting in a cost, when the target appeared in
19 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM an uncued region for the "mental spotlight" or attentional beam was already engaged at the cued location and had to be redirected to the new target-containing (uncued) location.
Features such as the size, processing distribution, and movement of the attentional beam have inspired a wealth of research in the area of attention. Each of these aspects will be discussed below. In terms of size, a large body of research has explored the idea that the attentional beam acts like the zoom lens of a camera and its focus can be contracted or expanded at will, depending on the task. Findings in support of this idea were first reported by Laberge (1983). In his experiment, participants were presented with five-letter words. In some trials, a probe appeared in the place of one of the five letters during the presentation of the word, or immediately following it. In an attempt to focus their attention, participants were asked to categorize either the central letter or the entire word. The idea was that categorizing the central letter would narrow the focus of the attentional beam, whereas categorizing the entire word would broaden its focus. In cases of the former, participants responded more rapidly to the probe when it appeared in the central letter position compared to any of the outer extreme positions. In cases of the latter, when participants were asked to categorize the entire word, reaction times did not vary to any significant extent based on the letter location of the probe. As predicted, reaction times were similar for all letter locations for the width of the attentional beam had been adjusted to accommodate processing of the entire word.
Another series of experiments testing the size of the attentional beam were conducted by Eriksen and St.-James (1986). They also reported findings in favour of the zoom lens model. Their results confirmed Laberge's (1983) initial finding that the width
20 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM of the attentional beam can be manipulated by precuing. In addition, they demonstrated that once the width of the attentional beam is adjusted in a narrow fashion, processing ability is more or less evenly distributed or homogeneous within the narrow surround. In contrast, their results also showed that when the width of the attentional beam expands, processing ability decreases for processing resources must then be dispersed within the newly-expanded beam. In terms of boundaries, their results suggested that there is no sharp demarcation in processing between the attentional beam and the rest of the visual field. Instead, it appears that beyond the attentional beam there is an inverse relationship between distance and processing ability. In other words, as distance from the boundary of the attentional beam increases, processing ability decreases in a graded dropoff-like manner.
Juola, Bowhuis, Cooper, and Warner (1991) investigated the matter of processing distribution within the context of the attentional beam as a zoom lens. Participants maintained a central fixation within a circle comprised of an inner, middle, and outer ring. A predictive cue then appeared in one of the three rings. This was followed by the onset of a target stimulus which also appeared within one of the three possible rings. The researchers hypothesized that if the attentional beam operates like a zoom lens, then performance should be significantly enhanced for targets appearing in the inner ring, irrespective of the ring in which the cue appeared for processing would be greatest at the foveal location. The results obtained were not as predicted. Juola and colleagues (1991) found that performance was greatest when the target appeared within the cued ring and that there was no significant effect of ring location. That is, performance could be
21 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM greatest in the middle or outer ring so long as the cue also appeared within that same ring.
These results suggest that processing is not necessarily greatest at the foveal location or homogenous within the attentional beam. Instead, it appears that attention can be allocated within the beam in an O-shaped pattern to only include the middle or outer ring which then results in the enhanced processing of these regions.
Adding to this line of evidence is another experiment conducted by Muller and
Hubner (2002). In this study, participants were presented with two superimposed letter streams. One stream consisted of large uppercase letters and the other of small uppercase letters which appeared superimposed in the centre of the larger letter (Figure 3).
Participants were instructed to maintain a central fixation and to attend to either the smaller central letter or to the larger background letter. Participants were also asked to detect the target letter H which could appear in the to-be-attended stream. Steady-state visual evoked potentials (SSVEPs), which exhibit greater amplitudes for attended versus unattended stimuli, were recorded from the participants' scalps. The SSVEPs only measured responses to the target letter in the small central letter stream. According to the researchers, if processing ability is homogeneous within the attentional beam, then
SSVEPs should be similar regardless of whether the small or large letter stream is attended, for attention should still remain contained within the boundary of the attentional beam. In other words, SSVEPs should not differ as a function of letter stream due to the spatial superimposition of the streams within the beam. The results reported by Muller and Hubnel (2002) are inconsistent with the hypothesis stated above. They found that
SSVEPs significantly increased when the small letters were attended versus ignored,
22 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Figure 3. Example of the two superimposed letter streams used in Muller and Hubner's (2002) study involving the attentional spotlight.
23 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM indicating that processing can differ for elements contained within the attentional beam depending on whether they are attended or ignored. Moreover, these results suggest that processing can be enhanced for the outer annular region of the attentional beam in a doughnut-like pattern when the center of the beam is ignored. The present findings, along with those above, provide converging evidence that processing within the attentional beam is not always contiguous or most powerful at the foveal location. Instead, it appears that attention can be allocated within the attentional beam in a variety of ways and that processing is therefore more flexible.
Another aspect of the attentional beam which has been investigated by researchers is the manner in which it moves within the visual field. According to Posner's spotlight analogy (Posner, 1980; Posner, Snyder, & Davidson, 1980; Posner & Cohen 1984;
Posner & Petersen, 1990), for the attentional beam to move within the visual field it must first be disengaged from its current spatial location, shifted to a new location, and then engaged or reallocated at that new location. The shift phase of this sequence has been the subject of much testing and research. Given that the attentional beam has been likened to a spotlight, one core assumption is there is a continuous and analog movement that occurs along its trajectory. As such, processing is facilitated for all illuminated spatial locations as they are traversed. Early findings reported by Tsal (1983) are consistent with this assumption. In his study, Tsal (1983) found that the attentional beam moved at a constant velocity over spatial locations and therefore took more time to travel longer distances. A subsequent study conducted by Remington and Pierce (1984) yielded an opposite pattern of results. These researchers found that the attentional beam moved just
24 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM as quickly to a target located two degrees from fixation, as it did to a target which was positioned ten degrees from fixation. In other words, the movement of the attentional beam showed no significant effects of distance. Others have also argued that there is insufficient empirical evidence to support previous models which propose that the attentional beam moves continuously through space (Yantis, 1998; Eriksen & Webb,
1989). Rather than illuminating all spatial locations along its trajectory, it appears as though the attentional beam jumps from one spatial location to the next and that processing resources are therefore engaged and disengaged in a more abrupt manner.
The Attentional Beam in Reading.
The role of attention as it relates to reading is perhaps best understood in terms of the E-Z Reader model (Reichle, Pollatsek, Fisher, & Rayner, 1998; Reichle, Rayner, &
Pollatsek, 2003; Reichle, Pollatsek, & Rayner, 2006). The E-Z reader model is a model of eye movement behaviour during reading. The major assumption underlying this model is that cognitive processes, and not oculomotor control, are essentially what drive eye movements during reading. Two core assumptions, present in the E-Z reader model, are that during normal reading 1) attention is allocated in a serial fashion on a word-by-word basis and 2) processes involving lexical access of the attended word are responsible for both saccadic programming (planning of eye movements) and the covert orienting of attention to the next word in the text.
The E-Z reader model describes the process reading in four main steps: 1) a familiarity check (early lexical processing), 2) the completion of lexical access (late lexical processing), 3) saccadic programming (planning of eye movements), and 4) the
25 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM execution of saccades (moving the eyes). Upon fixating a word, both a familiarity check and lexical access of the attended word begin to occur, and often the familiarity check is terminated first. The completion of this familiarity check (early lexical processing) is essentially what signals the programming of an eye movement to commence. The completion of lexical access (late lexical processing) proceeds in parallel with this saccadic programming. The termination of late lexical processing signals the reader's internal covert attention, or attentional beam, to shift to the next word in the text. This is then closely followed by an already-planned eye movement to the same location and the processing cycle begins anew. An important feature within this framework is that the shifting of the attentional beam described above precedes eye movements during normal reading. Additionally, processing begins immediately once the attentional beam is engaged on the next word in the parafovea, despite the fact the eyes remain on the previously attended word. Thus, in a sense, it is the shifting of the attentional beam and related parafoveal processing that propel eye movements to the next word in a text during reading.
Within the framework of eye movements and the attention beam, the E-Z reader model stipulates that the reader's attentional beam is shifted to the next to-be-fixated word while their eyes remain on the currently-fixated word. At this time, parafoveal information, including some orthographic and phonological properties, are extracted from the attended to-be-fixated word. This is then thought to facilitate the identification and processing of the subsequent word in a text (Rayner, 1999). A very robust finding with respect to parafoveal processing during the allocation of the attentional beam is that
26 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM fixation times on words which have been parafoveally processed are significantly shorter compared to words which are parafoveally masked on an earlier fixation (Morris, Rayner,
& Pollatsek, 1990; Rayner, 1975; Rayner, Well, Pollatsek, & Bertera, 1982; Schroyens,
Vitu, Brysbaert, & d'Ydewalle, 1999), thus reflecting a parafoveal or preview benefit.
Present Study
The concepts of attention and the attentional beam have been thoroughly reviewed above for each represents an inherent part of the Attentional Disengagement
(AD) model. Although research involving previous models is far more extensive, the AD model seems theoretically superior in its ability to explain the MLE across a variety of conditions (for example, see Roy-Charland et al., 2007 for response latency results).
There are still, however, important aspects of the model which require empirical testing.
In the present work, attempts will be made to test the pre-lexical attentional mechanism at the core of the AD model; the attentional beam. More precisely, the present thesis will provide the very first systematic exam of the attentional beam within the context of the missing-letter effect. Properties such as the size or scope, expansion, and processing homogeneity of the attentional beam will all be investigated. These ideas have never been empirically tested within the present context. Assuming attention does, indeed, act as a circular "mental spotlight" and that processing abilities are fairly consistent within the surround, the attentional beam depicted in the AD model will not only enhance processing of target-containing words, but of other elements or lexical features in the surround contained within the "mental spotlight". Similarly, disengaging attention and shifting the attentional beam to a new location, as is the case in normal reading, will, in
27 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM some cases, result in less efficient processing and detection of elements or features within previously engaged areas. In other words, the present hypothesis is that the MLE can be extended beyond letters to also include surrounding elements contained within the attentional beam. In line with this view, symbols will be incorporated in close proximity to words within a passage in an attempt to produce a missing-symbol effect (MSE).
Detection of a symbol in close proximity to a function word should result in higher rates of omission than detection of a symbol on a content word, for, in the case of the former, less attention is allocated to the region containing the function word. Contrary to other models of the MLE, the AD model predicts that omission rate for probes located in the space surrounding lexical items can vary as a function of lexical item characteristics. This would reflect the concept of the attentional beam. In this experiment, we will be using a novel and innovative strategy to test this hypothesis. Participants will be asked to read two separate texts for comprehension and to detect a series of minus signs (-) which will appear in both. Lines of text will be presented on a computer display screen and accompanied by rows of division signs (+), which will be positioned both above and below the text (Figure 4). Once the eyes cross an invisible boundary located on the last letter of a word preceding a critical word, the division sign (+) will be transformed into a minus sign (-) either 25 milliseconds (ms) or 170 ms later. Participants will then be expected to detect the minus sign (-).
As stated, two distinct texts will be used for the present purposes. The first, hereon referred to as the Des text, was used by Saint-Aubin et al. (2003) in testing the
Unitization Account/Processing Time Hypothesis and later by Roy-Charland et al.
28 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Invisible boundary _ i
Ti. , , Tou;- lc& .natiiio j 'jllui^ en ville pour rton boulot. J'otaxs ooiup table dans uno
[25msec after the eyes crossed the boundary] * -JSt-^ _t__ Tous les matins 3'allaio en villi pour men ioouiot. J'etair comptable dan° une |f
[30 msec later the display was back tonormal ] ] Cjjjp*^ „x_ _2 „ t Tous le- matins j 'allais en vill£ pour mon boulot. J'<§tais comptable dans une t
Figure 4. Example of method to be used in Experiment 1.
29 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
(2007; 2009) in testing the major assumptions underlying the GO model. The second text, hereon referred to as the Pour/cour text, was also used by Saint-Aubin et al. (2003) in testing the Unitization Account/Processing Time Hypothesis. These texts have been selected for this study for each has been used to empirically test previous models of the MLE. In other words, they will be used for the sake of comparison and in an effort to ensure a sound methodology. In addition, these texts are interesting in that they offer unique and respective control features for testing omission rates, many of which have been employed at one time or another within the history of the MLE field. Examples of these control features are discussed below.
The Des text will be used for it represents the equivalent of the English language
"the" text which has often been used in research involving the MLE. Most texts used in the field compare multiple occurrences of a frequent function word to a variety of control content words. Most importantly, the Des text will be used for it provides an opportunity to test a crucial aspect within the MLE field; word frequency. This text contains 24 occurrences of the word des (the), along with 24 occurrences of three-letter content words beginning with the letter "d". One reasonable objection or argument is that the content words in this text consist of different words and each possesses a different frequency whereas the function word des always remains the same. However, it is important to note that all content words are controlled for in terms of word length. Thus, the key feature within this text is that it enables researchers to isolate, manipulate, and explicitly test word frequency as a critical factor within the MLE. The importance of word frequency in
30 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM this field cannot be ignored for, although previous research has focused on word frequency and word role, models have tended to place a bigger emphasis or more weight on the idea of frequency over role in determining the MLE. Consider, for example, the use of "functors" as depicted in the GO model. The latter proposes that function words act as structural anchors within a sentence until the time attention is shifted to more meaning-laden elements within the sentence, such as content words. However, the use of function words as "functors" or structural anchors is, to a large extent, due to their high frequency. These words are processed more rapidly, thereby enabling their fast and early use within a sentence. Thus, in a very real sense, it is frequency which sets this process in motion. Overall, this text will provide an opportunity to isolate and test a sizable range of word frequencies due to its unique control feature. Furthermore, it will provide a novel opportunity to investigate the effect of different word frequencies within the context of the attentional beam.
The Des text is an obvious choice for this study for it represents a typical test utilized within the MLE field. Yet, given that the present objective is to explicitly test processing of the attentional beam, it is not sufficient on its own for it does not possess all of the necessary control features. In conjunction with the Pour/cour text, however, enough controls will be implemented, thereby enabling more comprehensive testing. The
Pour/cour text controls for the problem of having 24 instances of the same function word versus 24 instances of different content words for it contains 16 instances of both the function word pour (for) and the content word cour (yard). In addition, it possesses yet another unique and valuable control feature; the ability to regulate or manage the
31 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM influence of neighbouring words on eye movements during reading, or more importantly, on the attentional beam during reading. The idea that words preceding or following a critical word can influence eye movements has been extensively researched. For example, sometimes the processing of a word is not always completed by the time the eyes move to a new location in the text. This processing can, in turn, continue or "spill over" onto the next word, thereby influencing how long it is fixated. This phenomenon is hence known as the spillover effect (Rayner & Duffy, 1986; Rayner, Sereno, Morris,
Schmauder, & Clifton, 1989; Rayner, 1998). Another factor which influences eye movements during reading is the landing or launch site of the eyes on words. For instance, the launch site effect refers to the finding that the further the launch site is located on an engaged word from the center of the next to-be-fixated word, the further to the left the landing position will be located (McConkie, Kerr, Reddix, & Zola, 1988;
Rayner & Raney; 1996; Rayner, 1998). In this context, the Pour/cour text presents a distinct advantage for it was constructed in such a manner that each sentence with the function word "pour" has been matched with a sentence containing the content word cour. What is more, in each sentence within this pair, the word preceding or following the critical word (pour or cour) contains the same number of letters and belongs to the same syntactic category (function or content word). This unique characteristic allows researchers to isolate and explicitly test omission rates for function versus content words in the absence of any external influence on processing. Simply put, this design eliminates the possibility of confounding factors within the text because each occurrence of the word pour is matched in terms of context for each occurrence of the word cour. Thus,
32 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM any significant difference observed between the two must be attributed to the integrated effect of word frequency and role. The way in which this design will translate into the context of the attentional beam within the framework of the current study is fairly straightforward; explicit testing of the processing of the attentional beam will be achieved for the control features associated with the Pour/cour text will be such that the processing of words or regions preceding or following critical words will not influence the processing of the attentional beam. In addition, these control features will ensure that there is no occurrence of a spillover or launch site effect, as described above. Briefly, the control features of this text will allow greater control of the attentional beam, resulting in more efficient testing.
Hypotheses
Participants first presented with the Des text will be instructed to detect the appearance of a minus sign (-) situated above or below a line of text. Likewise, participants first presented with the Pour/cour text will be instructed to detect the same minus sign (-) whose position will also be counterbalanced above or below a line of text.
For half of the trials (both texts), the minus sign (-) will appear 25 milliseconds (ms) after the eyes cross the invisible boundary located on the word preceding the critical word.
This same timer will be set to 170 ms for the remaining half of the trials. These two probe delays were selected for a large body of research suggests that they represent contrasting situations in which attention is already allocated at fixation (25 ms) or possibly already disengaged and allocated to the next word (170 ms) (Rayner & Pollatsek, 1989; Fischer,
1999). According to the AD model, at 25 ms attention should still be engaged on both
33 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM content and function words and omission rates for the minus sign (-) should be similar. At
170 ms, there should be more omissions when the probe is located over a high-frequency function word than over a lower frequency content word because attention is likely to be disengaged from the high-frequency function word.
In the current study, the dependent variable will be the proportion of omissions on critical words in each experimental condition. This figure will be obtained by dividing the number of omissions by the number of occurrences within the text. The independent variables will consist of the probe delay (25ms or 170ms) and the test word type
(function or content). Proportions of omissions will be analyzed in a two-way repeated- measures analysis of variance (ANOVA) with probe delay (25ms or 170ms) and word type (function or content) each as within-subject factors.
In addition, the influence of word frequency on omission rate will be examined by analyzing the nine token content words in the Des text and computing item-based correlations. Specifically, these correlations will be based on the reported frequency of the content word and its observed proportion of omissions. The finding of a positive correlation between frequency and omission rate has been well documented. In the present study, however, no differential effect of frequency should be observed at 25ms for any of the content words for, according to the AD model and as discussed above, attention is still allocated in much the same manner as at the time of fixation. At 170 ms, on the other hand, an effect of frequency on omission rates should be observed within the nine token words for, as discussed previously, attention will likely disengage more rapidly from regions associated with higher frequency content words compared to lower
34 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM frequency content words. An effect of timing correlation between the two probe delay conditions is therefore expected.
Lastly, detection response latencies will be computed and analyzed for each participant. Response latencies represent the time elapsed from when the eyes cross the invisible boundary on the critical word until the time the participant makes a detection response by pressing the mouse button. Previous research involving the MLE has demonstrated that response latencies tend to be longer for higher versus lower frequency words, and for function versus content words (Roy-Charland et al., 2007; 2009). The AD model accounts for this finding by proposing that, because high-frequency function words result in a more rapid disengagement of attention than low-frequency content words, it is more difficult and time-consuming in cases of the former for the reader to accurately detect a target. Simply put, since attention is already disengaged, detection takes more time and response latencies are therefore longer. Consequently, longer response latencies should be observed for the detection of probes associated with regions containing high-frequency function words over those containing low-frequency content words. In terms of the probe delay experimental conditions, response latencies should be longer in the 170 ms versus the 25 ms condition for attention will likely have been disengaged from the target-containing region by 170 ms, resulting in a more delayed and less efficient detection strategy. At 25 ms, on the other hand, response latencies should decrease for attention should still be engaged in much the same manner as it is at the time of fixation. Simply put, attention will not have been disengaged at this point in time.
35 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
In addition, the influence of word frequency on reaction time (RT) will be examined by again analyzing the nine token content words in the Des text and computing item-based correlations. No differential effect of frequency on RTs should be observed at
25 ms for any of the content words for, according to the AD model, attention is still allocated in much the same manner as at the time of fixation. At 170 ms, on the other hand, a positive correlation between frequency and RT should be observed within the nine token words for attention will likely disengage more rapidly from regions associated with higher frequency content words compared to lower frequency content words. More precisely, RTs associated with higher frequency content words should be longer compared to those associated with lower frequency content words for, as stated, past research has demonstrated that RTs tend to be longer for higher versus lower frequency words, and for function versus content words (Roy-Charland et al., 2007; 2009).
Methodology
Participants
Participants were 32 undergraduate students recruited from Laurentian University.
French was the first language of all participants. In addition, all were required to have normal or corrected-to-normal vision in order to be eligible to participate in the study.
Material
The two experimental texts were presented to participants in black font and centered against a white display background. The Des text contained a total of 13 computer display pages. There were three to four lines of text per display page. The Des text also contained a total of 593 words (see Saint-Aubin et al., 2003; Saint-Aubin &
36 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Klein, 2004; Roy-Charland et al, 2007; 2009). There were 24 instances of the French plural indefinite article des (the), with a frequency count of 10,624.93 occurrences per million and 24 instances of nine different three-letter control content words beginning with the target letter "d". The average frequency count for the control content words was
404.07 occurrences per million (New, Pallier, Brysbaert, & Ferrand, 2004). These control words consisted of: don (gift or donation), dit (says), dix (ten), dos (back), duo (duet), dis
(says), dur (hard), due (due), and due (duke). The Pour/cour text contained a total of 18 computer display pages. There were three to four lines of text per display page. The
Pour/cour text also contained a total of 809 words (see Saint-Aubin et al., 2003). There were 16 instances of the French preposition pour (for) and 16 instances of the French noun cour (yard). The frequency count for the common wordpowr was 6198.24 occurrences per million, whereas it was only 150.14 occurrences per million for the more rare word cour (New et al., 2004). As discussed previously, each occurrence of pour and cour was preceded or followed by the same function or content word in the text.
Apparatus
Stimuli were presented on a 21-inch VIEW-Sonic CRT monitor. Eye movements were measured via infrared eye tracking using an SR Research Ltd. Eyelink II system
(SR 520 lens, monocular). This system has a high accuracy (<0.5 degrees) and a high sampling rate (500 Hz). The eye tracker device consists of two miniature cameras which are mounted onto a padded headband. Two cameras, located below the eyes of the participant, are used to measure the position of the eyes on the display screen every two ms. For the present purposes, these cameras allowed the easy selection of the
37 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM participant's dominant eye. An infrared sensor, which is integrated directly into the headband, is used for head-tracking which allows precise tracking of the participant's point of gaze. A distance of 60 centimeters (cm) between the participant and stimuli was maintained via a chin rest.
Procedure
The participant was seated in front of the computer display screen. Instructions were provided to the participant and a consent form was obtained. The eye tracker was placed on the participant's head and the binocular cameras below the eyes were adjusted accordingly. The calibration process then took place. Participants were asked to follow an icon (black circle with a white central point) which appeared at 13 respective locations on the computer display screen. Normally, a nine-point calibration is utilized in eye tracking experiments. However, due to the fact that the present task was entirely dependent on eye movements, a thirteen-point calibration was used. This process took place a second time in an effort to validate the original calibration cycle. In order to proceed, the total deviation between both calibration processes must be equal to or less than 0.5° of visual angle. Normally, a total deviation of 1° of visual angle is sufficient (see Roy-Charland et al, 2007); however, as mentioned above, the task at hand was wholly dependent on eye movements and therefore required greater precision. An eye contingent presentation procedure was used. To begin, participants were presented with either the Des or the
Pour/cour text and instructed to read the text for comprehension. In addition, they were instructed to detect the appearance of a minus sign (-). Above and below each line of text, a series of division signs (-^) were presented (Figure 4). Once the eyes cross an invisible
38 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM boundary situated on the last letter of a word preceding a critical word, a timer began. For half of the critical words in each text, the division sign (-^) was transformed into a minus sign (-) 25 ms after the eyes crossed the boundary. For the remaining half of the critical words in each text, the division sign (+•) was transformed into a minus sign (-) 170 ms after the eyes crossed the boundary. In all conditions, the division sign (+) was converted into a minus sign (-) by the deletion of the two dots. The minus sign (-) remained onscreen for 30 ms and then the two dots were represented anew. As stated, participants were instructed to detect the minus sign (-). This was achieved by pressing a mouse button as quickly and accurately as possible. Presentation order of the two experimental texts was completely counterbalanced across participants. There were a total of four randomized experimental conditions which were also counterbalanced across participants. The two probe delay experimental conditions (25 ms and 170 ms) were differentiated in the following manner for a total of four conditions; in half of the trials for each condition, the minus sign (-) appeared above the line of text, and in the remaining half it appeared below the line of text. For the Des text, there were a total of 54 probe transformations; 24 occurred above or below the letter d embedded in the frequent function word des, 24 occurred above or below the letter d embedded in the nine different rare content words, and 6 occurred above or below different letter positions in 6 distinct distracter words. For the Pour/cour text, there were a total of 38 probe transformations;
16 occurred above or below the letter r embedded in the frequent function word pour, 16 occurred above or below the letter r embedded in the rare content word cour, and 6 occurred above or below different letter positions in 6 distinct distracter words. At the
39 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM end of the experiment, participants were asked to answer 5 multiple-choice questions for each text (Appendices B & C) in an effort to ensure their full reading comprehension.
Results
The data involving the reading and the probe-detection task for both experimental texts were scored using the SR Research Eyelink II Data Viewer software. This program provides a visual overlay consisting of the participant's successive landing positions superimposed onto a corresponding passage of text. Additionally, the configuration of this software is such that the trigger (eyes crossing invisible boundary), probe transformation, and participant's button press response (detection) were all displayed as message events.
For all probe transformations, both detection and fixation status were recorded.
For detections, response latencies were computed by subtracting the time of the participant's button press response from the time the participants' eyes first crossed the invisible boundary on the last letter of the word preceding the target word. Importantly, response latencies were computed independent of fixation status. With respect to fixations, different measures were recorded, where applicable. These included first fixation time, single fixation time, gaze duration, and total fixation time. In line with previous research, fixations could occur either through first pass or through a regression
(Roy-Charland et al, 2007; 2009; Saint-Aubin & Klein, 2001). Moreover, only one landing position on a word was required for fixation. First, probabilities of fixation for both texts are presented below, followed by omissions, response latencies, and fixation durations.
40 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Probability of Fixation
For both experimental texts, performance level for the comprehension questions was evaluated. For the Des text, performance level on the comprehension questions
(Appendix B) was found to be 67.5% indicating that performance level was above chance and that participants were, in fact, reading for comprehension. Likewise, for the
Pour/cour text, performance level on the comprehension questions (Appendix C) was observed to be 75.6% indicating that, yet again, participants were reading for comprehension. Next, probabilities of fixation were computed by dividing the number of fixated target words by the total number of target word occurrences in each text.
Fixations were analyzed in a 2 X 2 repeated-measures analysis of variance (ANOVA) with word role (function and content) and probe delay (25 and 170 ms) as within-subject factors. In all analyses, a significance level of .05 was adopted. For the Des text, no main effects were observed for word role [F (1, 31) = 1.15,/>=29] or probe delay [F < 1]. In addition, no significant interaction emerged with respect to word role and probe delay [F
< 1]. Inspection of Table 1 and Figure 5 reveal no significant difference in terms of the probabilities of fixating a function or content word in both the 25 and 170 ms probe delay conditions. Results for the Pour/cour (Figure 6) text mimic those of the Des text. No main effects were found for word role [F(l, 31) = 1.27,;? = 27] or probe delay [F (1, 31)
= 2.25, p =.14], nor was a significant interaction [F (1, 31) = 2.18, p =.15] observed.
41 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Table 1 Means and Standard Deviations Probabilities of Fixation of the Target Region of the Des and Pour/cour Texts as a Function of Probe Delay and Word Role
Probe Delay
25 ms 170 ms
Fixation Fixation Target (N=32) (N=32)
Type M SD M SD
Des Function .65 .17 .65 .18 Content .69 .16 .68 .17 Fixation Fixation Target (N=32) (N=32)
Type M SD M SD
Pour/cour Function .86 .16 .79 .17 Content .80 .18 .80 .18
42 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Probabilities of Fixation - Des text 1 0.9
c 0.8 o IS 0.7 x * £ 0.6 ^•^ o .» *<*:?* 0,S J><: .2 DFunction | 0.4 (0 DContent | 03 * 0.2 0.1 0 25 ms 170 ms Probe Delay
Figure 5 Probabilities of fixation in the Des text.
Probabilities of Fixation - Pour/cour text
1 o, 9 c 0 .8 o ,7 « o x .6 X t o o .5 S o OFunction 4 V 1° ,.* OContent •§° 3 4P 4 0, 2
0. 1 V ' -CI 0 25 ms 170 ms Probe Delay
Figure 6 Probabilities of fixation in the Pour/cour text.
43 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Omissions
For both experimental texts, proportions of omissions were computed by dividing the number of omissions on probe transformations by the total number of occurrences in each text. For both texts, omissions were analyzed in a 2 X 2 repeated-measures ANOVA with word role (function and content) and probe delay (25 and 170 ms) as within-subject factors. Significant main effects were observed in the Des text for both word role [F (1,
31) = 7.60, rj2 =.20] and probe delay [F (1, 31) = 7.07, r\2 =.19], however no significant interaction was found [F < 1]. Inspection of Table 2 and Figure 7 reveal that proportions of omissions in the Des text were smaller for function words compared to content words in both the 25 ms and 170 ms probe delay conditions. In contrast to all previous research involving the missing-letter effect, participants omitted more probe transformations associated with content versus function words. Moreover, proportions of omissions in the
Des text were smaller in the 25 ms compared to 170 ms probe delay conditions for both function and content words, thereby indicating that participants were more accurate in the
25 ms compared to the 170 ms condition. In contrast to the Des text, no significant effects
(Figure 8) were reported in the Pour/cour text [all Fs < 1].
44 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Table 2 Means and Standard Deviations for Omissions of the Target Region of the Des and Pour/cour Texts as a Function of Probe Delay and Word Role
Probe Delay
25 ms 170 ms
Fixation Fixation Target (N=32) (N=32)
Type M SD M SD
Des Function .29 .20 .41 .22 Content .37 .25 .44 .20 Fixation Fixation Target (N=32) (N=32)
Type M SD M SD • —— —— — Pour/cour Function .38 .23 .39 .25 Content .39 .26 .44 .22
45 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Omissions - Des text 1 0.9 g 0.8 o •a 0.7 I 0.6 o 0.5 QFunction J 0.4 QContent 8.0.3 o £ 0.2 0.1 25 ms 170 ms 0 Probe Delay
Figure 7 Proportion of omissions in the Des text.
Omissions - Pour/cour text I 0.9 0.8 c o « 0.7 I/I | 0.6 o 0.5 QFunction | 0.4 QContent & 0.3 ' «-H {' o t0 £ 0.2 0.1 s 0 25 ms 170 ms Probe Delay
Figure 8 Proportion of omissions in the Pour/cour text.
46 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
For the sake of comparability, omissions for the Des text were then computed as a function of fixation status (Roy-Charland et al, 2007; 2009). Thus, the analyses were then based on a sample size of 29 participants, compared to the original 32 because of missing observations in some of the conditions. For this analysis, omissions were analyzed in a 2 X 2 X 2 repeated-measures ANOVA with word role (function and content), probe delay (25 and 170 ms), and fixation status (word fixated or skipped) as within-subject factors. These results replicated the significant main effect with respect to word role, as described above [F(l, 28) = 6.08, rj2 =.18]. Inspection of Table 3 and
Figure 9 reveal that proportions of omissions in the Des text were smaller for function compared to content words in both the 25 ms and 170 ms probe delay conditions.
Moreover, proportions of omissions were smaller for function versus content words in both the word fixated and word skipped fixation conditions. Overall, participants were more accurate in detecting probe transformations associated with function versus content words across all conditions. No main effects with respect to probe delay [F (1, 28) = 2.47, p =.13] or fixation status [F< 1] were observed, nor were any significant interactions found [all Fs < 1]. Omissions are not included here as a function of fixation status for the
Pour/cour text due to the fact the text contained less target words compared to the Des text, thereby resulting in a substantial number of missing observations and analyses based on a far more reduced number of participants.
47 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Table 3 Means and Standard Deviations for Omissions of the Target Region of the Des Text as a Function of Fixation Status, Probe Delay, and Word Role
Fixation Status
Fixated Skipped
25 ms 170 ms 25 ms 170 ms
Target (N=29) (N=29) (N=29) (N=29)
Type M SD M SD M SD M SD
Des Function .29 .20 .45 .30 .35 .35 .41 .32 Content .38 .31 .48 .23 .45 .32 .42 .30
48 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Omissions - Des text 1 c 0.9 g 'in 0.8 0.7 £ O 0.6 0.5 0.4 Q Function 0.3 0.2 OContent 0.1 0 25 ms | 170 ms 25 ms 170 ms
Fixated Skipped Probe Delay
Figure 9. Proportion of omissions in the Des text as a function of fixation status.
49 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Response Latencies
As mentioned, response latencies represent the time elapsed from the time of the probe transformation until the time the participant indicated a detection by means of a button press response. Response latencies were computed by averaging participants' response times to transformations in each distinct condition. The longest response latency confidently attributed to a target word was 2991ms in the Des text and 3577 ms in the
Pour/cow text.
For both texts, response latencies were analyzed in a 2 X 2 repeated measures
ANOVA with word role (function and content) and probe delay (25 and 170 ms) as within-subject factors. No significant main effects (Figure 10) were observed in the Des text for either word role [F < 1] or probe delay [F < 1], nor was a significant interaction found involving word role and probe delay [F (1, 30) = 2.12,p =.16]. In terms of the
Pour/cour text, no significant main effect was found for word role [F < 1]. However, a main effect emerged for probe delay [F (1, 30) = 36.02, rj2 =.55]. Inspection of Table 4 and Figure 11 reveal that the direction of this finding was such that response times appeared longer in the 25 compared to 170 ms condition. No significant interaction was noted for word role and probe delay [F < 1].
50 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Table 4 Means and Standard Deviations for Response Latencies of the Des and Pour/cour Texts as a Function of Probe Delay and Word Role
Probe Delay
25 ms 170 ms
Fixation Fixation (N=31) (N=31)
Type M SD M SD
Des Function 549 272 604 284 Content 623 162 568 137 Fixation Fixation Target (N=31) (N=31)
Type M SD M SD
Pour/cour Function 651 236 581 161 Content 650 164 586 145
51 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Response Latencies - Des text
1000 900 800 E 700 E 600 500 *M >%% DFunction
400 !• %$<£* OContent 300 :«. 200 100 ii 0 25 ms 170 ms Probe Delay
Figure 10 Response latencies in the Des text.
Response Latencies - Pour/cour text
1000 900 tn 800 "ST 700 E i= 600 ^1 500 DFunction 400 D Content 300 200 pis-*-' 100 if*
0 25 ms 170 ms Probe Delay
Figure 11 Response latencies in the Pour/cour text.
52 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Once again, for the sake of comparability, response latencies for the Des text were also computed as a function of fixation status (Roy-Charland et al., 2007; 2009). Due to missing observations, response latencies were based on 16 participants instead of the original 32 and analyzed in a 2 X 2 X 2 repeated-measures ANOVA with word role
(function and content), probe delay (25 and 170 ms), and fixation status (word fixated or skipped) as within-subject factors. Inspection of Table 5 and Figure 12 reveal that, as described in the above, no significant main effects were observed for either word role or probe delay [all Fs < 1]. Additionally, no significant effect emerged with respect to fixation status [F (1, 15) = 1.12,/? =.31], nor were any significant interactions found.
Once again, response latencies are not included here as a function of fixation status for the Pour/cour text due to a reduced number of target words and missing observations.
53 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Table 5 Means and Standard Deviations for Response Latencies of the Des Text as a Function of Fixation Status, Probe Delay, and Word Role
Fixation Status
Fixated Skipped
25 ms 170 ms 25 ms 170 ms
Target (N=16) (N=16) (N=16) (N=16)
Type M SD M SD M SD M SD
Des Function AAA 573 553 143 602 229 551 196 Content 568 123 577 149 595 182 560 222
54 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Response Latencies - Des text 1000 - 900 '- 800 - _ 700 < | 600 * « 500 >- tl' E 400 - sit, H 300 > "4 p v Q Function 200 L D Content 100 ~ i 0 - m yt 25 ms | 170 ms 25 ms 170 ms
Fixated Skipped Probe Delay
Figure 12 Response latencies in the Des text as a function of fixation status.
55 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Fixation Durations
Eye movement measures of fixation duration were recorded for each participant across all conditions. Consistent with previous literature (Rayner, 1998; Rayner et al.,
2004; Roy-Charland et al., 2007; 2009), four main measures were applied to the data: first-fixation duration, single-fixation duration, gaze duration, and total fixation time.
First-fixation duration represents the time spent on a single fixation or the first in a series of subsequent fixations. Single-fixation duration consists of the time spent on a word when only one forward fixation occurs. Gaze duration represents the sum of all forward fixations on a word that occur before a saccade to another word takes place. Finally, total fixation time represents the sum of all fixations, including those occurring on the first pass or through a regression.
For the sake of comparability (Roy-Charland et al., 2007; 2009), fixation duration measures for the Des text were initially analyzed in a 2 X 2 X 2 repeated-measures
AN OVA with word role (function and content), probe delay (25 and 170 ms), and detection status (transformation detected or omitted) as within-subject factors. No main effects were found with respect to word role. However, main effects for probe delay were observed across three conditions. These included first fixation duration, single fixation duration, and gaze duration [F(l, 11) = 6.01, rj2=.35; F(l, 9) = 6.52, r;2 = 42; F(l, 11)
= 6.82, r\2 =.38]. In line with the results presented above with respect to rates of omission and response latencies, inspection of Table 6 reveals that, in addition to spending more response time and being more accurate, fixation duration measures were significantly longer in the 25 versus 170 ms probe delay condition. In addition, main effects for
56 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM detection status were observed across three conditions. These included first fixation duration, and gaze duration, and total fixation duration [F (1, 11) = 10.69, rj2 =.49; F(l,
11) = 7.81, t]2=A2; F(l, 14) = 14.29, r}2 = 51]. Not surprisingly, inspection of Table 6 also reveals that all fixation duration measures were longer for detections versus omissions, except for function words in the gaze condition. However, due to the fact that detection status was collapsed across word role, gaze durations were still longer overall for detections versus omissions. The totality of these results supports previous findings by Roy-Charland et al. (2007) concerning fixation duration and detection status.
57 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Table 6 Means and Standard Deviations for Fixation Duration Measures of the Des Text as a Function of Detection Status, Probe Delay, and Word Role
Probe Delay
25 ms 170 ms
First Single Total First Single Total Fixation Fixation Gaze Fixation Fixation Fixation Gaze Fixation Duration Duration Duration Duration Duration Duration Duration Duration (N=12) (N=10) (N=12) (N=15) (N=12) (N=10) (N=12) (N=15)
M SD M SD M SD M SD M SD M SD M SD M SD
Omission Function 294 124 307 109 327 182 321 164 277 151 303 161 316 146 362 123 Content 325 155 315 109 331 153 341 140 275 156 248 98 293 157 362 204
Detection Function 417 135 370 137 447 143 475 189 288 100 311 102 297 98 336 132 Content 425 226 396 250 437 224 435 134 350 122 278 95 361 120 435 172
58 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
While these results certainly provided some insight as to the significant relationships which exist between probe delay and detection status, it was observed that these analyses were once again based on a reduced number of participants due to missing observations. The remaining analyses for both the Des and Pour/cour text were therefore collapsed across word role. In other words, all remaining fixation duration measures data were analyzed in a 2 X 2 repeated-measures ANOVA with probe delay (25 and 170 ms) and detection status (transformation detected or omitted) as within-subject factors
(Appendix A). Importantly, all findings presented below were based on a larger number of participants much closer to the original N value.
For the Des text, main effects for probe delay were again observed across three fixation duration measures. These included first fixation duration, single fixation duration, and gaze duration [F (1, 27) = 9.51, r\2 = 26; F(l, 24) = 3.31, tj2 =. 12; F(l, 27)
= 10.73, rj2 =.28]. Inspection of Table 7 and Figure 13 reveal that, once again, the direction of these findings was such that participants fixate longer in the 25 versus the
170 ms condition. For the Pour/cour text, no main effects emerged for probe delay on any fixation duration measure. In terms of detection status, main effects were observed across all fixation duration measures in the Des text; first fixation duration, single fixation duration, gaze duration, and total fixation duration [ F(l, 27) = 38.35, r\2 =.59; F
(1, 24) = 21.04, r\2 =.47; F(l, 27) = 31.65, r\2 =.54; F(l, 28) = 19.27, ^2= 41].
Moreover, main effects for detection were also observed across all four fixation duration measures in the Pour/cour text [F(\, 25) = 40.47, f}2 = 62; F (1, 21) = 29.62, r\2 =.59; F
(1, 25) = 39.81, rj2 = 61; F(l, 27) = 44.34, ri 2 =.62]. Inspection of Table 7 and Figure 14
59 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM reveal that, in line with previous research (Roy-Charland et al., 2007), fixation duration measures were longer for detections versus omissions on all measures in both texts.
Interactions did not reach significance on any condition in both texts.
60 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Table 7 Means and Standard Deviations for Fixation Duration Measures of the Des and Pour/cour Texts as a Function of Detection Status and Probe Delay
Probe Delay
25 ms 170 ms
First Sir igle Total First Single Total Fixation Fixation Gaze Fixation Fixation Fixation Gaze Fixation Duration Duration Duration Duration Duration Duration Duration Duration (N=25) (N=25) (N=28) (N=29) (N=25) (N=25) (N=28) (N=29)
M SD M SD M SD M SD M SD M SD M SD M SD
Des text Omission 297 87 301 98 321 102 346 121 247 75 268 106 270 79 327 95 Detection 383 110 385 112 413 120 440 121 334 85 338 110 352 95 389 113
(N=26) (N=22) (N=26) (N=28) (N=26) (N=22) (N=26) (N=28)
Pour/cour text Omission 274 70 287 82 315 102 397 123 252 118 292 141 294 122 354 132 Detection 413 122 538 201 499 132 567 180 413 166 431 223 494 250 552 285
61 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Fixation Duration Measures - Des text
QOmission Q Detection
25 ms Probe Delay
Figure 13. Fixation duration measures in the Des text.
Fixation Duration Measures - Pour/cour text
DOmission D Detection
First [Single | Gaze First Single Gaze
25 ms 170 ms Probe Delay
Figure 14. Fixation duration measures in the Pour/cour text.
62 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Word Frequency
The Des text contained 24 instances of des, with a frequency count of 10,624.93 occurrences per million and 24 instances of nine different three-letter control content words beginning with the target letter "d". Importantly, these three-letter control words were of varying frequencies, the average of which was 404.07 occurrences per million
(New et al., 2004). Item-based correlations were computed on the Des text data set to test the influence of word frequency on rates of omission and response latencies with respect to the probe transformations associated with the 9 different three-letter control content words. These analyses used the normal log of frequency due to the fact the frequency values varied between 2 and 2,602 occurrences per million (New et al., 2004). In the 25 ms probe delay condition, no significant correlations emerged for either log frequency and rate of omission or log frequency and response latencies, respectively [r = .29, p
=.16; r = .06, p =.79]. Likewise, in the 170 ms probe delay condition, there were also no significant findings with respect to either log frequency and rate of omission or log frequency and response latencies, respectively [r = -.05,p =.83; r = .37,p =.08].
Discussion
Overview
On the whole, these results did not produce a typical or usual missing-letter effect.
Instead, and quite surprisingly, they revealed a significant reverse missing-letter effect.
To put it differently, a differential rate of omission was observed for probes associated with function versus content words in the Des text; inspection of Tables 2 and 3 revealed a higher rate of omission for probes associated with content versus function words.
63 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Importantly, these results represent the very first of their kind within the entirety of the very robust missing-letter effect field. The implications of this unexpected finding are further discussed below.
Perhaps as a result of this, the present findings also did not replicate or extend previous research involving probabilities of fixation, response latencies, or word frequency. First, as can be seen in Table 1, no differences emerged in either text with respect to the likelihood of fixating a function over a content word; likewise, there were no differences in terms of the probability of fixating a word in the 25 ms versus 170 ms probe delay condition. Previous research, on the other hand, has demonstrated that fixations are more likely to occur on content versus function words when a standard missing-letter effect is produced (Roy-Charland et al., 2007). This makes intuitive sense for function words are more closely associated with omissions and content words with detections. Therefore, one reasonable assumption is that fixations would occur more frequently in connection with detections, and thus, probes associated with content words.
Second, as shown in Tables 4 and 5, response latencies to the probe did not significantly vary as a function of word role in either text. It is worth noting that previous research has consistently found response latencies to be longer for function compared to content words (Roy-Charland et al., 2007; 2009; Saint-Aubin et al., 2003).
In terms of the AD model, this is explained by the idea that disengagement of the attentional beam on function words occurs more rapidly compared to content words.
This, in turn, results in limited time to access deeper and more complex knowledge of the word, such as information pertaining to the word's constituent letters. Longer response
64 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM latencies are therefore observed when a detection occurs. Again, the fact the present findings did not yield this same pattern of results may be relevant to the direction of the missing-letter effect. Response times to the probe were significantly different across the probe delay conditions, but only in the Pour/cour text. More precisely, response latencies were longer in the 25 ms compared to the 170 ms probe delay condition. This finding is also explored in greater depth below.
Furthermore, the present results did not reveal an effect of word frequency on either omission rate or response latencies. Item-based correlations were computed on the nine token content words in the Des text. In past research when a standard missing-letter effect has been observed, the finding of a positive correlation between frequency and omission rate and frequency and detection response latencies have been well documented
(Roy-Charland et al., 2007). Once again, the fact that a reverse missing-letter effect is observed may be in part responsible for this discrepancy. What is more, one possibility is that certain mechanisms involved in the standard missing-letter effect, such as frequency, do not play a critical role or do so to a lesser extent when the direction of the effect is reversed.
In terms of eye movement measures of fixation duration, our results are again inconsistent with previous research demonstrating that fixation durations tend to be shorter on function versus content words (Roy-Charland et al., 2007), for as evidenced in
Table 6, no significant differences emerged with respect to word role in either text. Here again, the direction of the missing-letter effect could be underlying this issue. By the same token, as can be seen in Tables 6 and 7, our results from both texts perfectly mirror
65 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM previous findings that fixation durations tend to be longer for detections versus omissions
(Roy-Charland et al., 2007; 2009). In fact, fixation durations were longer for detections compared to omissions in both texts and across all four measures. In 2009, Roy-Charland and colleagues provided the first empirical evidence that the longer fixation durations connected with detections are comprised of both the processing of the word and the planning and production of a motor (detection) response. Additionally, it was observed in the present study, and as shown in Table 6, that the first fixation duration, single fixation duration, and gaze duration were all significantly longer in the 25 ms compared to the
170 ms probe delay condition in the Des text only. It is worth noting that these three measures are conditional on the target word being fixated in the initial pass and are therefore collectively referred to as forward fixation measures. On the contrary, regressions are comprised of all fixations, whether forward fixations or regressions.
Together with the aforementioned results involving omissions and response latencies, it appears that in some circumstances, participants fixated target words for longer periods, took longer to respond to probe transformations, and were more accurate at detecting probe transformations in the 25 ms versus the 170 ms probe delay condition.
Most importantly and as discussed above, the present study failed to produce a standard missing-letter effect; instead a rather unexpected reverse missing-letter effect emerged.
These findings are discussed below within the architecture of the original hypotheses.
Reverse Missing-Letter Effect
It was originally predicted that at 25 ms there would be no differential rate of omission for probes associated with either function or content words. On the other hand,
66 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM it was hypothesized that a higher rate of omission would arise for probes associated with high-frequency function versus low-frequency content words in the 170 ms probe delay condition. This would be due to the fact that research suggests these two probe delays represent contrasting situations in which attention is already allocated at fixation (25 ms) or possibly already disengaged and allocated to the next word (170 ms) (Rayner &
Pollatsek, 1989; Fischer, 1999). Thus, at 170ms, attention should be disengaged from regions associated with function compared to content words. As stated, the results did not confirm the original hypothesis. Instead, they revealed an opposite pattern of results and, quite shockingly, a reverse missing-letter effect was observed.
Different possibilities were investigated to account for this rather unexpected finding. First, a methodological limitation could, in theory, be responsible for this result.
Recalling that probe transformations took place either above or below a target function or content word, it could be argued that the scope of processing associated with the attentional beam of the AD model was not sufficiently large enough to account for the detection of probe transformations. Otherwise put, transformations were occurring in regions outside of the scope of the attentional beam and cognitive processing was therefore not homogenous, as originally predicted, for lexical and non-lexical features, such as the probe.
Second, according to Rayner (2009), during reading, location of available information can be differentiated into three regions: the foveal region (represented the central 2 degrees of fixation), the parafoveal region (representing the central 5 degrees of fixation), and the peripheral region (representing the total region beyond the parafovea).
67 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Moreover, it has been well established that readers extract meaningful information in the forward parafoveal region while fixating the previous word (Inhoff & Rayner, 1986;
McConkie & Rayner, 1975; Rayner, 1975; 2009). Interestingly, recent research has also investigated the scope of backward parafoveal or post-view processing in readers and determined that useful information can also be extracted by readers to the left of fixation
(Binder, Pollatsek, & Rayner, 1999; Roy-Charland, Saint-Aubin, Lalande, Belanger, &
Klein, 2011; Wang, Tsai, Inhoff, & Tzeng, 2009). Thus, it appears that a parafoveal preview benefit exists for readers both to the right and left of fixation. Within the context of the current results, even if the scope of processing of the attentional beam was sufficient to account for the detection of probe transformations, it could also be argued that, in reading, a parafoveal preview benefit does not exist above or below target words in the same straightforward manner as the well-documented forward and even backward parafoveal preview benefits.
Several lines of research (Haikio, Bertram, Hyona, & Niemi, 2009; Rayner,
1986; Rayner, 2009) support the idea that a reader's perceptual span, and thus, the extent to which parafoveal information is processed in the attentional beam, varies as a function of reading level and skill. Hence, the aforementioned explanation could be further supported by the following idea; parafoveal preview benefits also evolve in readers over time not only as a function of skill, but as a function of their respective reading direction.
To put it differently, if perceptual span can be modulated by reading skill, then presumably, parafoveal preview may also be modulated by reading direction. In this vein, due to the inherent nature of left-to-right reading for instance, vertical parafoveal preview
68 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM would not necessarily imply a benefit for the reader, or would do so to a much lesser extent than traditional horizontal parafoveal preview. Again, within the context of the present findings, this could serve to explain the absence of a usual missing-letter effect for the original hypothesis regarding omission rate was contingent upon homogenous processing for both the text and probe within the beam.
Generally speaking, both of these possibilities seem somewhat plausible in accounting for the absence of a standard missing-letter effect. However, neither one on its own can account for the presence of a reverse-missing letter effect for this implies that a differential rate of attentional disengagement was nevertheless still occurring, albeit in the opposite direction. With this in mind, the ideas that 1) parafoveal processing is facilitating probe detections and 2) the scope of processing of the attentional beam is sufficient in accommodating for probe transformations, cannot be altogether refuted.
Yet another possibility with respect to the present findings involves the main mechanisms of the Structural Account of the missing-letter effect (Koriat and Greenberg,
1994). According to this model, the first step in normal reading consists of the extraction of sentence structure by the reader. In this process, function words serve as structural anchors and remain in the cognitive forefront until a sentential frame or skeleton has been constructed. Next, these function words recede to the cognitive background and the reader's attention is shifted to semantically-rich content words in the sentence, whereby meaning is then derived. In the context of the missing-letter effect, target letters located within function words (structural anchors) quickly becomes less accessible for constituent letter processing and their subsequent detection, because of their unique cognitive
69 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM dissipation described above. Otherwise put, omissions are due to the diminished perceptual status of structural anchors. Content words, on the other hand, remain relatively stable once they are propelled to the cognitive forefront, for semantic analysis represents the final step in text processing.
The present finding of a reverse missing-letter effect could be explained in terms of the idea that at both 25 and 170ms, function words still linger in the cognitive forefront, for sentence structure has not yet been established. Consequently, lexical information, including information pertaining to the word's constituent letters, is readily available to the reader, thereby facilitating detection. By the same token, the transition from structure to meaning has not yet been triggered and fundamental lexical information pertaining to content words, including letter information, is not yet accessible to the reader. As a result, letter or probe detection is impaired.
For the most part, the reverse missing-letter effect fits nicely within this line of reasoning. Nevertheless, some issues still remain unresolved. For instance, this explanation fails to account for the important role of the attentional beam within the context of reading and the missing-letter effect. Moreover, the fact that participants fixated target words for longer periods, took longer to respond to probe transformations, and were more accurate at detecting probe transformations in the 25 versus 170 ms condition suggests that a differential rate of attentional disengagement was indeed occurring at these times.
All things considered, we therefore advance what we believe to be the most parsimonious interpretation of the reverse missing-letter effect: a dynamic attentional
70 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM beam that varies as a function of the cognitive load. A large body of research has investigated the idea that the attentional beam is dynamic and acts like the zoom lens of a camera in that it can contract or expand at will, depending on the task (Eriksen & St-
James, 1986; Juola, Bowhuis, Cooper, & Warner, 1991; Laberge, 1983). Similarly, several lines of research, especially research in connection to the E-Z reader model, support the idea that normal serial reading processes are under the control of an attentional beam (Rayner, 1998; 2009; Reichle, Pollatsek, Fisher, & Rayner, 1998;
Reichle, Rayner, & Pollatsek, 2003; Reichle, Pollatsek, & Rayner, 2006). What remains unanswered, however, is the precise manner in which mechanisms or features of the attentional beam, such as size and elasticity, operate in normal reading. Barring studies which examined the attentional beam in relation to isolated words (Laberge, 1983) or superimposed letter streams (Muller & Hubner, 2002), no research pertaining to the elasticity of the attentional beam exists in normal reading tasks.
Within the context of our current results, we propose that, during the processing of normal text, differing degrees of cognitive resources are required for the processing of function and content words. According to our interpretation, cognitive processing of target words begins in the parafovea while fixating the previous word (Rayner, 2009).
Because of their often highly familiar visual configuration and syntactic role, we assume that function words represent a lighter cognitive load than do lower-frequency content words. This, in turn, accelerates the cognitive processing of function words after initial fixation has occurred. One possibility is that cognitive load also modulates the elasticity of the attentional beam before the disengagement of the beam or covert shift takes places
71 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM in the direction of the next to-be-fixated word. More precisely, we propose that cognitive processing is terminated at an earlier stage for function compared to content words which, in turn, signals the attentional beam to expand in a concentric manner. As a result of this expansion, the attentional breadth of the beam becomes sufficiently large enough to accommodate the probe transformations, thereby facilitating detection. In contrast, due to their semantically rich nature, content words represent a much heavier cognitive load which, in turn, delays the expansion the attentional beam. In this case, we propose that the signal to disengage the attentional beam or shift covert attention to the next to-be- fixated word precedes the expansion signal. Thus, the beam remains rather narrowly focused during processing and detections associated with content words are impaired, for probe transformations are occurring outside of the boundaries of the attentional beam.
At first glance, this explanation seems inconsistent with earlier research involving the AD model of the missing-letter effect. In 2010, Saint-Aubin, Klein, Deacon, and
Thompson (2010) successfully produced a missing-colour effect in readers. That is to say, readers "missed" more target colours associated with function versus content words, as per a standard missing-letter effect. Importantly, these findings not only provided empirical support for the concept of an attentional beam within the context of the MLE, they also suggested that processing was homogenous for both lexical and non-lexical features within the beam. Our study differs in that, although our probes represented non- lexical features, they are also not directly letter or lexically-bound, as were the colours.
Applying our interpretation to these results, it could be argued that a typical missing- letter effect emerged due to the fact the task did not require concentric expansion of the
72 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM attentional beam. Instead, it involved detection of a specific target colour of a letter which was presented entirely within the scope of the beam upon initial fixation of the word.
Thus, one reasonable assumption here is that cognitive processing was terminated at an earlier stage for function compared to content words which, in turn, signaled the disengagement of the attentional beam, altogether bypassing the expansion signal. The end result was that participants omitted more target colours associated with function words. Applying these same principles to content words, it folloyvs that cognitive processing occurred for a longer period of time, thereby facilitating the detection of target colours.
Probe Delay Conditions
One further point worth noting here was that all original hypotheses and predictions were contingent upon the idea that the mechanisms of attention operated differently at 25 and 170 ms. More specifically, it was assumed that at 25 ms attention was, in all likelihood, still engaged from the time of original fixation, whereas at 170 ms attention was likely disengaged and shifted to the next word in the text (Rayner &
Pollatsek, 1989; Fischer, 1999). Moreover, it was hypothesized that at 170 ms, omissions would be greater for probes associated with function versus content words for attention would not necessarily have disengaged from regions containing content words by this time, thereby facilitating probe detections.
As stated, the results revealed that participants fixated target words for longer periods, took longer to respond to probe transformations, and were more accurate at detecting probe transformations in the 25 versus 170 ms probe delay condition. This
73 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM could reflect that fact that at 25 ms, attention is still engaged on regions containing both function and content words, whereas it is wholly disengaged from both at 170 ms. In terms of the former, estimates involving the speed of attention shifts typically range from
25 -75ms (Egeth & Yantis, 1997; Inhoff, Eiter, & Radach, 2005; Inhoff, Eiter, & Radach,
2006). The EZ reader model, for instance, proposes that there is a 50 ms "eye-to-mind" transmission delay during which time attentional shifts cannot occur for processing of the fixated word is taking place (Pollatsek, Rayner, & Reichle, 2006). This finding is largely based on estimates of the time needed for covert attentional shifts to occur from one location to the next, as reported in the classic study by Treisman and Gelade (1980) and also by Posner (1980). Simply put, the EZ reader proposes that attention remains engaged on a fixated word for a minimum of 50 ms for this also represents the approximate time required for an attentional shift to take place. For these reasons, our probe delay of 25 ms likely fell within the hypothesized range of initial attentional engagement and produced the aforementioned results.
In terms of the latter, recent research conducted in the area of parafoveal processing (Inhoff, Eiter, & Radach, 2005; Inhoff, Eiter, & Radach, 2006) suggests that the disengagement of attention may occur much earlier than 170 ms. For instance, the extraction of useful linguistic information from a parafoveal target was found to begin between 70-140 ms after the onset of pretarget viewing and to continue beyond that time.
Within the present research context, one reasonable assumption is that the disengagement process must be initiated or taking place in order for this extraction of parafoveal information to occur. Thus, our disengagement timing estimate of 170 ms may be
74 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM inaccurate. In fact, at this time, one possibility is that parafoveal processing could even be taking place beyond the next word at the second to-be-fixated word (n+2). If our probe delay of 170 ms actually exceeds the hypothesized range of attentional disengagement
(70-140 ms), then no differential rate of omission for target regions containing function and content words should be occurring. Additionally, fixation duration measures should be shorter on the basis that attention has already disengaged from the target region. What is more, evidence from previous research (Roy-Charland et al., 2007; 2009) has suggested that, in line with the current results, shorter fixations are more closely associated with omissions, which appear to characterize the 170 ms probe delay condition. In short, reducing the probe delay to a value in closer proximity to the upper boundary of the hypothesized range of attentional disengagement, namely 140 ms, might be effective in producing the desired differential rate of omission for probes associated with function versus content words.
Limitations
The current findings argue against a typical missing-letter effect for the results yielded an opposite pattern of results and generated a reverse missing-letter effect. It has been proposed this finding is due to the malleability of the attentional beam which, to an extent, is under the control of the cognitive load associated with a word. What must be noted here, however, is the reverse missing-letter effect was not observed in the
Pour/cour text. This is problematic for, although the Des text represents a typical and widely-used text in the missing-letter effect field, it also does not possess the unique control features which are found in the Pour/cour text. Examples of these include the fact
75 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM the Pour/cour text controls for the problem of having 24 instances of the same function word versus 24 instances of different content words for it contains 16 instances of both the function word pour (for) and the content word cour (yard). In addition, the Pour/cour text presents a distinct advantage for it was constructed in such a manner that each sentence with the function word pour has been matched with a sentence containing the content word cour. In each sentence within this pair, the word preceding or following the critical word (pour or cour) contains the same number of letters and belongs to the same syntactic category (function or content word). This eliminates the possibility of confounding factors within the text because each occurrence of the word pour is matched in terms of context for each occurrence of the word cour. Moreover, these control features ensure there are no occurrences of either a spillover or launch site effect accounting for the results. Overall, the control features of the Pour/cour text would allow greater control of the attentional beam and perhaps provide a more solid foundation upon which to build our theory involving the dynamic attentional beam, cognitive load, and the missing-letter effect.
The fact that significant results only emerged in the Des text is also problematic due to the fact that the Des text only contains the frequent function word des and nine different rare content words. In this vein, it could be argued that participants were always more accurate in detecting probe transformations associated with the word des because 1) they were not reading for comprehension and 2) they were using the word des as a cue in detecting the probe transformation. The idea that this task does not tap into normal reading cognitive processes and, instead, is a mere signal detection task is further
76 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM supported by the following observations: 1) a typical missing-letter effect was not produced, 2) word frequency did not appear to modulate either omissions or response latencies with respect to the nine different content words in the Des text, 3)- no significant differences emerged with respect to the likelihood of fixating a content over a function word, and 4) no significant differences emerged with respect to fixation duration measures involving function and content words. At first glance, these findings appear to support the signal detection task perspective. However, as mentioned, the performance level on the comprehension questions for the Des text was above chance indicating that participants were, in fact, reading for comprehension. Secondly, although word frequency, probability of fixation, and fixation duration measures for word type did not reach significance, it must be mentioned that all were in the same direction as those found in traditional reading and/or missing-letter effect tasks. Thirdly, it could also be argued here that if participants were, in fact, using the function word des as a cue due to its high frequency, then a reasonable assumption here is they would also use the frequent function word pour and perhaps even the rare content word cour as similar cues in detecting the probe due to their high frequency in the Pour/cour text. This would, in turn, decrease rates of omission for both word types. To put it differently, participants would be equally accurate in detecting probe transformations occurring in regions containing both function and content words. In comparison to the Des text, however, inspection of Table 2 and
Figure 8 do not reveal higher levels of accuracy in the Pour/cour text in detecting the probe transformation. In fact, rates of omission appear to be fairly similar to those observed in the Des text.
77 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
An additional clarification which must be made involves the homogeneity of processing within the attentional beam. All original hypotheses and predictions were contingent upon the idea that processing distribution should be homogenous for both words and symbols contained within the surround of the attentional beam. However, several lines of research suggest that processing is not always uniformly allocated within the beam. Eriksen and St.-James (1986), for instance, proposed that an inverse relationship exists between distance from the central fixation and processing ability.
Another study conducted by Juola, Bowhuis, Cooper, and Warner (1991) found that processing distribution could vary within the beam as a function of cuing location.
Finally, according to Muller and Hubner (2002), processing ability is not always contiguous or most powerful at the foveal location; instead their findings suggest that, to an extent, the allocation of covert attention to specific regions within the beam is essentially what governs processing ability. All things considered, it must therefore be noted that differential processing of words and probes in the surround may have been occurring and influencing our pattern of results. Although we do not altogether refute this possibility, it could also be argued here that the nature of the probe transformation was such that it did not require extensive perceptual processing for probe transformations to minus signs simply consisted of the temporary deletion of the superior and inferior dots of the division signs. Thus, we remain confident this straightforward change could be successfully detected even in the face of diminished perceptual processing.
78 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Conclusion /Future Direction
In summary, the present study demonstrated a differential rate of omission for function versus content words; however, this effect did not emerge in the expected direction. Instead, a reverse missing-letter effect was observed. To put it differently, a higher rate of omission was observed for probes associated with content versus function words. For these reasons, we proposed that the scope of processing of the attentional beam varies as a function of the cognitive load associated with a word. More precisely, function words represent a lighter cognitive load than do content words which accelerates processing. What is more, fewer cognitive resources are recruited for processing which also, in turn, permits the expansion of the attentional beam. This facilitates the detection of probes associated with function words. In terms of content words, the attentional beam remains more narrowly-restricted and focused due to the fact content words represent a heavier cognitive load and more resources are therefore recruited for their processing.
Thus, detections are impaired, for probe transformations are nevertheless occurring outside of the beam's boundaries. Although these findings offer preliminary support for the idea of a dynamic attentional beam during reading, future studies should first be directed towards reproducing the reverse missing-letter effect in another text. As stated, it could be argued that, in the Des text, the frequent function word des is acting as a cue and participants are therefore more accurate in detecting the probe transformation. By the same token, it could further be argued that content words in this same text cannot act as cues due to the fact that they are varied throughout the text. In order to circumvent this issue, another text could be constructed which not only varies its content words, but
79 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM function words as well. This, in turn, would eliminate the possibility that participants were using a cue to detect the probe transformation. Furthermore, this could also lend support to the idea that participants are, indeed, reading for comprehension and not simply detecting a probe transformation. Other future studies within the present context should be aimed at explicitly and systematically testing this theory of a dynamic attentional beam by manipulating the cognitive load associated with a word.
Alternatively, future studies could also be designed to manipulate probe distance from the foveal region to explicitly test the elasticity or expansion properties of the attentional beam in relation to cognitive load.
In a broad sense, reading can be defined as the complex orchestration of many different cognitive processes. We acknowledge that much in terms of the precise cognitive mechanisms underpinning reading remain largely unknown. What has been revealed here, however, is that certain lexical properties may be modulating the elasticity of the attentional beam and, to a greater extent than previously thought, influencing cognitive processing within the context of reading.
80 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
References
Binder, K. S., Pollatsek, A., & Rayner, K. (1999). Extraction of information to the left of
the fixated word in reading. Journal of Experimental Psychology: Human
Perception and Performance , 25, 1162-1172.
Corcoran, D. (1966). An acoustic factor in letter cancellation. Nature, 210, 658.
Crundall, D., Cole, G. G., & Galpin, A. (2007). Object-based attention is mediated by the
collinearity of targets. The Quarterly Journal of Experimental Psychology, 60
(1), 137-153.
Davis, G., & Driver, J. (1997). Spreading of visual attention to modally versus amodally
completed regions. Psychological Science , 8 (4), 275-281.
Desimone, R. (1998). Visual attention mediated by biased competition in extrastriate
visual cortex. Philosophical Transactions of the Royal Society B: Biological
Sciences, 353 (1373), 1245-1255.
Desimone, R., & Duncan, J. (1995). Neural mechanisms of selective visual attention.
Annual Review of Neuroscience, 18, 193-222.
Duncan, J. (1984). Selective attention and the organization of visual information. Journal
of Experimental Psychology: General, 113 (4), 501-517.
Egeth, H. E., & Yantis, S. (1997). Visual attention: Control, representation, and time
course. Annual Review of Psychology, 48, 269-297.
81 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Eriksen, C. W., & St.-James, J. D. (1986). Visual attention within and around the field of
focal attention: A zoom lens model. Perception and Psychophysics, 40, 225-240.
Eriksen, C. W., & Webb, J. M. (1989). Shifting of attentional focus within and about a
visual display. Perception & Psychophysics , 45 (2), 175-183.
Fisher, M. H. (1999). An investigation of attention allocation during sequential eye
movement tasks. The Quarterly Journal of Experimental Psychology, 52A (3),
649-677.
Greenberg, S. N., & Koriat, A. (1991). The missing-letter effect for common function
words depends on their linguistic function in the phrase. Journal of Experimental
Psychology Learning, Memory, & Cognition, 17, 1051-1061.
Greenberg, S. N., Healy, A. F., Koriat, A., & Kreiner, H. (2004). The GO model: A
reconsideration of the role of structural units in guiding and organizing text on
line. Psychonomic Bulletin and Review, 11 (3), 428-433.
Haikio, T., Bertram, R., Hyona, J., & Niemi, P. (2009). Development of the letter identity
span in reading; Evidence from the eye movement moving window paradigm.
Journal of Experimental Child Psychology, 102, 167-181.
Healy, A. F. (1976). Detection errors on the word the: Evidence for reading units larger
than letters. Journal of Experimental Psychology: Human Perception and
Performance , 2 (2), 235-242.
82 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Healy, A. F. (1994). Letter detection: A window to Unitization and other cognitive
processes in reading text. Psychonomic Bulletin and Review, 1 (3), 333-344.
Huey, E. (1908). The Psychology and Pedagogy of Reading. New York: Macmillan.
Inhoff, A. W., & Rayner, K. (1986). Parafoveal word processing during eye fixations in
reading; Effects of word frequency. Perception and Psychophysics , 40, 431-439.
Inhoff, A. W., Eiter, B. M., & Radach, R. R. (2006). Temporal overlap in the linguistic
processing of successive words in reading: Reply to Pollatsek, Reichle, and
Rayner (2006a). Journal of Experimental Psychology: Human Perception and
Performance, 32(6), 1490-1495.
Inhoff, A. W., Eiter, B. M., & Radach, R. (2005). Time course of linguistic information
extraction from consecutive words during eye fixations in reading. Journal of
Experimental Psychology: Human Perception and Performance, 31 (5), 979-995.
James, W. (1890). The Principles of Psychology. New York: Henry Holt.
Juola, J. F., Bowhuis, D. G., Cooper, E. E., & Warner, C. B. (1991). Control of attention
around the fovea. Journal of Experimental Psychology: Human Perception and
Performance, 75,315-330.
Koriat, A., & Greenberg, S. N. (1994). The extraction of phrase structure during reading:
Evidence from letter detection errors. Psychonomic Bulletin and Review, 1 (3),
345-356.
83 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Laberge, D. (1983). The spatial extent of attention to letters and words. Journal of
Experimental Psychology: Human Perception and Performance , 9, 371-379.
McConkie, G. W., & Rayner, K. (1975). The span of the effective stimulus during a
fixation in reading. Perception andpsychophysics, 17, 578-586.
McConkie, G. W., Kerr, P. W., Reddix, M. D., & Zola, D. (1988). Eye movement control
during reading: The location of initial eye fixations on words. Vision Research,
28(10), 1107-1118.
Moravcsik, J. E., & Healy, A. F. Effect of meaning on letter detection. Journal of
Experimental Psychology: Learning, Memory, & Cognition, 21, 82-95.
Morris, R., Rayner, K., & PoUatsek, A. (1990). Eye movement guidance in reading: The
role of parafoveal letter and space information. Journal of Experimental
Psychology: Human Perception and Performance, 16, 268-281.
Muller, M., & Hubner, R. (2002). Can the spotlight of attention be shaped like a
doughnut? Evidence from steady-state visual evoked potentials. Psychological
Science, 13(2), 119-124.
New, B., Pallier, C, Brysbaert, M., & Ferrand, L. (2004). Lexique 2: A new French
lexical database. Behavior Research Methods, Instruments, & Computers, 36,
516-524.
PoUatsek, A., Reichle, E. D., & Rayner, K. (2006). Serial processing is consistent with
the time course of linguistic information extraction from consecutive words
84 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
during eye fixations in reading: A response to Inhoff, Eiter, and Radach. Journal
of Experimental Psychology: Human Perception and Performance , 32 (6), 1485-
1489.
Posner, M. I. (1980). Orienting of attention. Quarterly Journal of Experimental
Psychology, 32, 3-25.
Posner, M. I., & Cohen, Y. (1984). Components of visual orienting. In H. Bouma, & D.
Bouwhuis (Eds.), Attention and Performance X(pp. 531-556). London: Lawrence
Erlbaum Associates Ltd.
Posner, M. I., & Petersen, S. E. (1990). The attention system of the human brain. Annual
Review ofNeuroscience , 13, 25-42.
Posner, M. I., Snyder, C. R., & Davidson, B. J. (1980). Attention and the detection of
signals. Journal of Experimental Psychology: General, 109 (2), 160-174.
Rayner, K. (1986). Eye movements and the perceptual span in beginning and skilled
readers. Journal of Experimental Child Psychology, 41, 211-236.
Rayner, K. (1998). Eye movements in reading and information processing: 20 years of
research. Psychological Bulletin, 124 (3), 372-422.
Rayner, K. (1975). Parafoveal identification during a fixation in reading. Acta
Psychologica, 39, 271-282.
85 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Rayner, K. (2009). The 35th Sir Frederick Bartlett Lecture: Eye movements and attention
in reading, scene perception, and visual search. Quarterly Journal of
Experimental Psychology, 62, 1457-1506.
Rayner, K. (1999). What Have We Learned About Eye Movements During Reading? In
R. Klein, & P. McMullen, Converging Methods for Understanding Reading and
Dyslexia (pp. 23-56). MIT Press.
Rayner, K., & Duffy, S. A. (1986). Lexical complexity and fixation times in reading:
Effects of word frequency, verb complexity, and lexical ambiguity. Memory &
Cognition, 14(3), 191-201.
Rayner, K., & Pollatsek, A. (1989). The Psychology of Reading. Englewood Cliffs, NJ:
Prentice-Hall.
Rayner, K., & Raney, G. F. (1996). Eye movement control in reading and visual search:
Effects of word frequency. Psychonomic Bulletin and Review, 3, 238-244.
Rayner, K., Ashby, J., Pollatsek, A., & Reichle, D. (2004). The effects of frequency and
predictability on eye fixations in reading: Implications for the E-Z reader model.
Journal of Experimental Psychology: Human Perception and Performance, 20,
720-732.
Rayner, K., Sereno, S. C, Morris, R. K., Schmauder, A. R., & Clifton, C. J. (1989). Eye
movements and online comprehension processes. Language and Cognitive
Processes, 4, 21-49.
86 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Rayner, K., Well, A., Pollatsek, A., & Bertera, J. (1982). The availability of useful
information to the right of fixation in reading. Perception & Psychophysics ,31,
537-550.
Reichle, E. D., Pollatsek, A., & Rayner, K. (2006). E-Z reader: A cognitive-control,
serial-attention model of eye-movement behavior during reading. Cognitive
Systems Research , 7, 4-22.
Reichle, E. D., Pollatsek, A., Fisher, D. L., & Rayner, K. (1998). Toward a model of eye
movement control in reading. Psychological Review, 105 (1), 125-157.
Reichle, E. D., Rayner, K., & Pollatsek, A. (2003). The E-Z reader model of eye
movement control in reading: Comparisons to other models. Behavioral and
Brain Sciences, 26 (4), 445-476.
Reichle, E., Warren, T., & McConnell, K. (2009). Using E-Z Reader to model the effects
of higher-level language processing on eye movements during reading.
Psychonomic Bulletin Review , 1-21.
Remington, R., & Pierce, L. (1984). Moving attention: Evidence for time-invariant shifts
of visual selective attention. Perception & Psychophysics, 35 (4), 393-399.
Roy-Charland, A., & Saint-Aubin, J. (2006). The interaction of word frequency and word
class: A test of the GO model's account of the missing-letter effect. The Quarterly
Journal of Experimental Psychology, 59 (1), 38-45.
87 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Roy-Charland, A., Saint-Aubin, J., Klein, R. M., & Lawrence, M. (2007). Eye
movements as direct tests of the GO model for the missing-letter effect.
Perception andPsychophysics, 69 (3), 324-337.Roy-Charland, A., Saint-Aubin,
J., Lalande, A., Belanger, A., & Klein, R. (2011). Eye movements when reading:
The importance of the word to the left of fixation. Manuscript submitted for
publication.
Roy-Charland, A., St.-Aubin, J., Lawrence, M., & Klein, R. (2009). Solving the chicken
and egg problem of letter detection and fixation duration during reading.
Attention, Perception, & Psychophysics, 71, 1553-1562.
Saint-Aubin, J., & Poirier, M. (1997). The influence of word function in the missing-
letter effect: Further evidence from French. Memory and Cognition, 25, 666-676.
Saint-Aubin, J., Roy-Charland, A., & Klein, R. M. (2003). Direct assessments of the
processing time hypothesis for the missing-letter effect. Journal of Experimental
Psychology : Human Perception and Performance , 29 (6), 1191-1210.
Schroyens, W., Vitu, F., Brysbaert, M., & d'Ydewalle, G. (1999). Eye Movement control
during reading: Foveal load and parafoveal processing. Quarterly Journal of
Experimental Psychology, 52A, 1021 -1046.
Treisman, A., & Gelade, G. (1980). A feature-integration theory of attention. Cognitive
Psychology, 12,97-136.
88 AD Model, MLE, Attentional Beam
THE AD MODEL OF THE MLE: A TEST OF THE ATTENTIONAL BEAM
Tsal, Y. (1983). Movements of attention across the visual field. Journal of Experimental
Psychology: Human Perception and Performance , 9, 523-530.
Wang, C, Tsai, J., Inhoff, A. W., & Tzeng, O. J. (2009). Acquisition of Linguistic
Information to the left of fixation during the reading of Chinese text. Language
and Cognitive Processes, 24, 1097-1123.
89 Appendix A - Analyses of Variance for Fixation Duration Measures
First Single Gaze Total
Df df df MSE Source df MSE n\ MSE 1P MSE n\ n\
Pes text
Probe Delay 1,27 9.51* 7099 .26 1,24 3.31 12223 .12 1,27 10.73* 8025 .28 1,28 2.54 14084 .08
Detection status 1,27 38.35* 5479 .59 1,24 21.04* 7061 .47 1,27 31.65* 6732 .54 1,28 19.27* 9270 .41
PDXDS 1,27 0.001 6511 .00 1,24 .18 7239 .01 1,27 0.08 9701 .003 1,28 0.78 9407 .03
Pour/cour text
Probe Delay 1=25 0.25 12891 .01 1,21 2.43 23746 .10 1,25 0.17 24003 .01 1,27 1.24 18667 .04
Detection status 1-25 40.47* 14378 .62 1,21 29.62* 28347 .59 1,25 39.81* 24036 .61 1,27 44.34* 21372 .62
PDXDS 1,25 0.34 9046 .01 1,21 2.96 23218 .12 1,25 0.10 18545 .004 1,27 0.19 28295 .01 Appendix B - Des Questions de comprehension
1.) Quelle sorte de patisserie Ginette s'est-elle achetee?
A) Francaise B) Anglaise C) Beige
2.) Quelle sorte de medicaments Ginette consomme-t-elle?
A) Prozak B) Valium C) Aspirine
3.) Depuis combien longtemps Mike est-il en prison?
A) Quelques semaines B) Quelques jours C) Quelques mois
4.) Selon Ginette, qui est responsable de remprisonnement de Mike?
A) Lejuge B) Lasociete C) Le gouvernement
5.) Selon Ginette, qui est responsable de la criminalite des enfants?
A) Leur pere B) Les voisins C) Leurs amis Appendix C - Pour/cour Questions de comprehension
1) Quel emploi occupe le narrateur de cette histoire ?
A) Vendeur d'autos usagees B) Conducteur d'autobus C) Comptable dans une banque
2) Qu'est-ce que le narrateur a recu comme cadeau de Noel de son travail ?
A) Un flamant rose B) Une fontaine en plastique C) Une boite a outils
3) Pourquoi les humanoides ont-ils choisi de se poser dans la cour du narrateur ?
A) Parce qu'ils voulaient utiliser la piscine B) Parce qu'ils voulaient leur rendre visite C) Parce qu'ils trouvaient la decoration de la cour tres belle
4) Combien de yeux les humano'ides avaient-ils ?
A) Deux yeux B) Trois yeux C) Quatre yeux
5) A quel endroit les humanoides se sont-ils poses dans la cour du narrateur ?
A) A cote de la piscine B) A cote de la fontaine en plastique C) A cote du garage