When the WRENCH turns a few heads: Expectation and semantic relatedness in inattentional blindness

Semantic relatedness and expectation were investigated in inattentional blindness — failure to perceive an unexpected object in plain sight when attention is engaged elsewhere. Participants named primary-task pictures and ignored distractor pictures. Four trials preceded a ‘critical ’ trial where an unexpected six-letter-word appeared at fixation, simultaneously with the pictures. In Experiment 1, we found robust effects for both in-lab and on-line-Zoom methodology. More participants reported the unexpected word semantically-related to the primary-task pictures than a semantically-unrelated word. In Experiment 2, expectations were violated, by changing the semantic category of the primary-task pictures. More participants reported the unexpected word semantically-related to the unexpected picture category than a semantically-unrelated word. When attentional resources are consumed by a task, a violation to task expectations is not enough to reorient attention to an unexpected word. Attention reorients to what is meaningful to the task, and what is meaningful is updated in light of unexpected information.


Introduction
What we perceive is influenced by more than objective reality.As we go about our waking life, the visual system receives light as physical input, converting it into electrical signals that are sent to the brain.Research on attention reveals that human perception is much more than this as it is derived from an interplay between external signals and internal processes.White and Aimola Davies began their 2008 paper stating that humans believe themselves "infallible witnesses to the realities of the external world" (p.1092).However, the highly researched phenomenon of inattentional blindness (IB), describing the failure to perceive something immediately in front of a person's eyes when their attention is engaged elsewhere (Mack & Rock, 1998), demonstrates that human perception is not always an accurate reflection of the external environment.
The exploration of the attentional system does not only have theoretical importance but it also has practical implications.Current research highlights the importance of identifying the factors involved in attention, so that we may learn how to mitigate significant and potentially devastating perceptual errors.One striking example comes from a study conducted by White and O'Hare (2022), in which novice pilots, who had undergone training as participants for this experiment, were asked to conduct a flight simulation inside a custom-made cabin.While distracted, these participants failed to report a rare and visually distinctive Boeing-787 aircraft passing within their field of vision.During one phase of the flight, participants followed a flight path with no distractions.Then in a second phase, the experimenter simultaneously engaged participants in a distracting telephone conversation.After the task, the participants completed a questionnaire regarding their detection of certain objects during the flight.Despite the importance of these objects to the task, during the distraction phase, participants failed to detect critical objects within their visual field-demonstrating that attentionally-demanding tasks can lead to IB for important and potentially dangerous information within a simulated real-world context.Another striking example comes from a study conducted by Simons and Schlosser (2017), in which police academy trainees and experienced police officers participated in a simulated vehicle traffic incident, where they approached a vehicle to issue a warning or citation to the driver.During this interaction, the driver's behaviour varied between passive and cooperative or agitated and hostile.Despite the task's real-world relevance and potential threat, 58 % of trainees and 33 % of experienced officers failed to report a visible gun on the passenger's dashboard-demonstrating that even those trained to be vigilant to potential threat can experience IB for a potentially dangerous object in a naturalistic real-world context.

Inattentional blindness
The study of the limits on attention is not new, as it has its roots in early investigations of selective attention (Neisser & Becklen, 1975).Neisser and Becklen's participants were tasked with following the action of a basketball game in one video recording while ignoring a handball game in another.The findings from two different experiment conditions-binocular and dichoptic-were a clear demonstration of the limitations of simultaneous attention to multiple stimuli.Although participants were able to attend to one video recording and ignore another, they had difficulty with noticing "odd events", such as a mid-game handshake or the disappearance of the ball, in the unattended recording.Continuing on with this seminal work, a series of experiments by Simons and Chabris (1999) demonstrated that perception can fail when attentional resources are consumed by a demanding visual task.Participants were instructed to count the number of basketball passes between players on a given team (wearing either black or white shirts) playing a passing game on a video recording.Of the 192 participants tested across 16 different conditions, a substantial proportion (46 %) failed to report an unexpected critical stimulus-a person wearing a black gorilla suit walking across the computer screen through the centre of the game.Simons and Chabris concluded that their results provided evidence in support of Mack and Rock's (1998) claim that "… attention is essential for conscious perception.Without it, we simply do not perceive" (p.74).In this experiment, the participant's attention is focussed on the specific goal of counting basketball passes.The focussing of attention based on internal factors such as goals, expectation, and prior knowledge, has been attributed to the dorsal attentional network (dorsal frontoparietal network) and is associated with 'top-down' attention (i.e., orienting to stimuli important to an internal goal, such as completing a task; Corbetta et al., 2008).The failure to report the unexpected person wearing the gorilla suit is an example of IB resulting from these top-down attentional processes.On the other hand, the automatic capture of attention by salient or unexpected stimuli in the environment has been attributed to the ventral attentional network (ventral frontoparietal network) and is associated with 'bottom-up' attention (i.e., detecting stimuli that are salient, and that unexpectedly capture one's attention; Anticevic et al., 2010).If detection of the unexpected person wearing the gorilla suit were to occur, this would be an example of bottom-up attentional capture.Therefore top-down and bottom-up attentional processes combine to determine the deployment of attentional resources and subsequently, what we perceive.
Inattentional blindness is a robust effect that has been replicated using a myriad of experiment paradigms (Redlich et al., 2021a), which in recent years has also included online data collection methods.Traditionally in IB research, participants complete an experiment in person in a laboratory setting where the experimenters have a high degree of control over the experiment set-up.Whereas for online methods, participants complete an experiment on their personal computers at home where the experimenters have less control over the experiment set-up.Much of this online data collection has been conducted using object-tracking dynamic paradigms (see Most et al., 2001;2005) in which stimuli such as shapes or letters move around a computer display and bounce off the edges of the display.Participants count the number of times stimuli of one kind (e.g., squares) bounce off the edges of the display while ignoring other stimuli (e.g., circles).On the critical trial, an unexpected object moves through the display.There have been several replications with this dynamic paradigm online (Stothart et al., 2015;Ward & Scholl, 2015;Wood & Simons, 2017).Other online IB studies have used static rather than dynamic paradigms (see Mack & Rock, 1998).These static paradigms involved static displays of stimuli in which participants were instructed to engage in a primary task, such as judging which of two arms of a cross was longer or monitoring strings of letters to count the number of 'real' words.On the critical trial, an unexpected object simultaneously appeared in the display next to the primary-task stimuli.There have been several adaptations of the static paradigm online, finding similar robust effects (Kreitz et al., 2015;Redlich et al., 2021b;Wood & Simons, 2019).Online IB studies demonstrate that there is minimal influence from variables such as screen type, size, or device on which the task is performed.This shift in conducting experiments-from the traditional laboratory setting to online methodologies-provides an opportunity to examine IB in diverse and ecologically-relevant contexts, enhancing the external validity and generalizability of findings.

Meaning and expectation
The occurrence of IB can be mitigated by the presence of stimuli that hold meaning and relevance to the observer.For example, Simons and Chabris (1999) found that despite participants failing to report a person in a black gorilla suit during a demanding task of counting basketball passes, attention was more likely to reorient toward the unexpected event (the gorilla) when the tracked team's shirts matched the colour of the gorilla suit.A review by Redlich et al., (2021a) indicates that certain stimuli with personal significance, such as an individual's own name (Mack & Rock, 1998), can reorient attention.However, it is noteworthy that this response did not extend to names per se (i.e., the names of other individuals), suggesting that this phenomenon is related to the attentional priority associated with meaningful cues.Moreover, researchers in IB have found higher rates of noticing for other meaningful stimuli, including faces (Devue et al., 2009), animate stimuli (such as animals; Calvillo & Jackson, 2014), body parts (Downing et al., 2004), emotionally-salient pictures (New & German, 2015) and items semantically related to a primary task (Koivisto & Revonsuo, 2007).Koivisto and Revonsuo (2007) investigated whether an unexpected stimulus, a critical word, could escape IB when it was semantically related to the primary-task picture category in a picture-naming task.The task involved attending to and naming pictures from either the category animals or furniture.For example, pictures of animals were presented within a white circular disk alongside pictures from other categories (i.e., distractor pictures).In the third and fourth trial, a critical word appeared in the centre of the computer screen concurrently with the primary-task pictures.Participants were asked "Did you notice anything new or additional that was not present in the previous trials?"and "If you did, what and where?" (Koivisto & Revonsuo,p. 847).The results demonstrated that participants were more likely to report the critical word (e.g., KISSA, cat in Finnish) semantically related to the primary-task picture category (e.g., animals) than an unrelated critical word.Koivisto and Revonsuo termed this the 'semantic-congruency effect', where attention reorients away from the current task (e.g., naming pictures from the category, animals) toward the critical word that is semantically related to, or congruent with, internal goals (in this example, the category to-be-named).Thus, it seems that despite the robust effect of IB, particular stimuli, predefined by varying internal priorities, can reorient attention away from the current task.Rock et al. (1992) described the methodology of IB as either the "elimination of attention or elimination of expectation" (p.532).Expectation, defined as the perceived probability of an event (Summerfield & Egnar, 2009), plays a significant role in attentional processes.Graziano (2014;2022) proposed an attentional schema, which is continuously describing the current state of attention and making predictions for the purpose of controlling attention.These predictions are influenced by monitoring the external environment and updating the existing attentional model based on previous experiences and expectations.Studies of attention in medical professionals provide supporting evidence for the role of expectation in attention (Drew et al., 2013;Williams et al., 2021).A study conducted by Drew et al. (2013) revealed that when radiologists undertook the familiar task of searching for lung nodules in patient scans, they missed an unexpected salient image of a gorilla embedded in the scans.The radiologists' attention was primarily directed toward the lung nodules due to their task-related expectations.In a separate study by Williams et al. (2021), radiologists did not report a clinically meaningful 9.1 cm breast mass that was unrelated to their task of detecting lung nodules.Despite the clinical importance of the breast mass, the task of finding lung nodules may have narrowed the radiologists' attention, causing them to miss this unexpected stimulus.Taken together, these studies show that task-related expectations can lead to instances of IB.Specifically, when attention is engaged elsewhere, Drew et al. underscored how an unexpected yet meaningless stimulus can be missed, whereas Williams et al. highlighted that an unexpected meaningful stimulus can also be missed if it is not related to the participant's task.
The violation of a held expectation has been shown to disrupt attentional focus, prompting a shift in attention to an unexpected event or stimulus and thus reducing the occurrence of IB.For example, White and Aimola Davies (2008) disrupted participants' expectations in an IB task-they violated task expectations by making the number of stimuli presented inconsistent with task instructions.Participants were tasked with naming letters, and all participants were presented with two letters on the arms of a cross.However, participants were either validly-cued (cue indicated two letters would be presented), over-cued (cue indicated three letters would be presented) or under-cued (cue indicated one letter would be presented).When participants' numerical expectations were violated, either by over-cueing or by under-cueing, the rates of IB for an unexpected square were significantly reduced compared to when participants' numerical expectations were consistent with the number of letters presented.These findings illustrate that not only can expectations cause people to miss something visually salient in the environment they are searching when expectations of the task are met but also that when expectations of the task are not met, the rates of IB can be significantly reduced.Together, these findings indicate the important role of expectation in modulating the deployment of attention.

Present study
The present study used an inattentional blindness paradigm and included a series of two experiments, with a total of eight experiment conditions tested.Each experiment consisted of seven trials, where participants attended and named primary-task pictures for a given semantic category while ignoring distractor pictures from a different semantic category.In the first four Non-Critical Trials, only the expected categories of pictures appeared on the computer display.In Trials 5, 6 and 7, in addition to the pictures, an unexpected word appeared at fixation, which will be referred to as the critical word from now on.Report of this critical word in Critical Trial 5 was of key research interest.The semantic categories of the primary-task pictures and critical words were manipulated to investigate the role of semantic relatedness on attention as it pertains both to expectation and to the violation of expectation.
To the best of our knowledge, this study is the first time that an exemplar-naming paradigm has been conducted using online methodology.For Experiment 1, 50 % of participants were tested in a traditional experiment laboratory setting and the other 50 % of participants were tested online using the video-conferencing platform Zoom.The reason for using two different data collection methods is that in-person data collection was halted in August 2021 due to a state-wide lockdown in response to the COVID-19 pandemic.This methodology for Experiment 1 allowed for a direct comparison between the findings obtained using traditional inlaboratory data collection methods and online data collection methods.One important distinction to emphasise here is between the experiment set-up in our Experiment 1 and previous online methodology.For Experiment 1, the experimenter was always present, either in-person for the in-laboratory data collection method or on Zoom for the online data collection method.For Experiment 2, 90 % of participants were tested online using the video-conferencing platform Zoom.

Experiment 1 (Conditions 1a, 1b, 1c, 1d)-semantic relatedness
Experiment 1 was a conceptual replication of the research conducted by Koivisto and Revonsuo (2007).We aimed specifically to replicate their findings of a semantic-congruency effect-a critical word semantically related to the primary-task picture category S. Chu and A. Aimola Davies would be reported more than a semantically-unrelated critical word.We used the primary-task picture categories of insects and tools (alongside distractor pictures of fruits).The critical word presented in Critical Trial 5-simultaneously with the pictures-was either semantically related or semantically unrelated to the category of primary-task pictures.Depending on the experiment condition, our critical words were either exemplars of the attended primary-task-picture category (semantically-related) or they were semantically unrelated to the primary-task picture category, but they were never exemplars of the distractor-picture category that participants were instructed to ignore.We predicted that in Experiment 1, although all participants were presented with a critical word, a higher proportion of participants would report a critical word semantically related to the primary-task picture category than a critical word semantically unrelated to the primary-task picture category..3.2. Experiment 2 (Conditions 2a,2b,2c,2d)

-violation of expectations
The appearance of the critical word amongst pictures in Experiment 1 could be considered a violation of the task-expectation for pictures, and Koivisto and Revonsuo (2007) found that with this set-up it was the semantically-related critical word that was reported by participants and not the semantically-unrelated critical word.But an unexpected word may not in and of itself directly interfere with the participants' task of looking for and naming insects (or tools).White and Aimola Davies (2008) more directly manipulated participants' primary-task expectations with a violation to their numerical expectation, which resulted in the report of the unexpected stimulus.
No study, to the best of our knowledge, has looked at both directly violating expectations and the semantic-congruency effect within the same IB paradigm.Therefore, in Experiment 2, we aimed to investigate the effects of directly violating participants' expectations regarding the primary-task picture category, and the interaction between expectation and semantic relatedness on noticing of a critical word.In Critical Trial 5, participants in all conditions experienced a violation to their task expectations for the primary-task picture category because the pictures in the instructed category were unexpectedly changed from insects to tools (or musical instruments).The critical word presented in Critical Trial 5 was either semantically related or semantically unrelated to this new and unexpected picture category.We predicted that in Experiment 2, although all participants were presented with an unexpected picture category, a higher proportion of participants would report a critical word semantically related to the unexpected picture category than a critical word semantically unrelated to the unexpected picture category.

Participants
There were 209 neurologically healthy participants recruited from the Australian National University (ANU) and general population via flyers and online sources.A total of 192 participants (131 females, 60 males, 1 other) remained in the study for statistical analysis (M age = 29.54,SD age = 7.56, range age : 18-49) 1 after excluding 17 participants who did not meet the inclusion criteria.Participants were excluded if English was not their most fluent language (total = 10; Condition 1a = 1, 1b = 2, 1c = 1, 1d = 1, Condition 2a = 3, 2c = 2) and if their primary-task accuracy for Trials 1 to 4 was below the 62.5 % cut-off (total = 7; Condition 1c = 2, Condition 2a = 1, 2c = 3, 2d = 1).Data was not excluded from 29 participants (Condition 1a = 2, Condition 1b = 6, Condition 1c = 1, Condition 1d = 2, Condition 2b = 9, Condition 2c = 4, Condition 2d = 5) who failed to detect the critical word in the Full-Attention Trial.In previous studies, the Full-Attention Trial (where participants no longer engaged in the primary task and instead were asked to look just at the trial display) was used to determine whether the critical word was reported when attentional resources were available.In those studies, participants who failed to report the critical word in the Full-Attention Trial were sometimes excluded.However, it has since been shown that excluding participants based on this criterion may understate the pervasiveness of IB (White et al., 2018).Data was also not excluded from 57 participants (Condition 1a = 8, Condition 1b = 10, Condition 1c = 7, Condition 1d = 5, Condition 2a = 6, Condition 2b = 3, Condition 2c = 7, Condition 2d = 11) who had indicated prior knowledge of IB.This decision was based on the findings of Beanland and Pammer (2010), who reported no significant effect of participants' previous knowledge of IB on rates of noticing of a critical stimulus in an IB paradigm. 2 The 192 participants included for statistical analysis had normal or corrected-to-normal vision, no history of neurological disorder, and were right-handed (M = 69.44;SD = 25.46) as assessed by the Edinburgh Handedness Inventory (Oldfield, 1971).These participants were randomly assigned to one of eight experiment conditions, with 24 participants in each condition: Experiment 1 = four 1 The present study included a wide range of participant ages.Previous research indicates that susceptibility to IB may increase with age (Graham & Burke, 2011;Horwood & Beanland, 2016;Stothart et al., 2015).Stothart et al. found that an increase of 10 years of age was associated with a 1.3fold increase in the probability of IB; therefore, in the present study, participant age was distributed equally across the eight experiment conditions.
2 To investigate the conclusions from Beanland and Pammer (2010), who stated "that possessing minimal knowledge of IB research does not influence the likelihood of experiencing IB in an experimental setting" (p.14), we ran an analysis comparing the proportion of noticers who reported they had knowledge of IB and the proportion of noticers who reported they did not.The results of the 2 'Knowledge of IB' (knowledge, no knowledge) x 2 'Noticers' (noticers, non-noticers) chi-square test across the 8 experiment conditions indicated that there was no significant difference between the proportion of noticers who reported they had knowledge of IB and those who reported they did not (p > 0.59).There were also no significant differences found when the experiment conditions were tested in pairs, for the semantically-related conditions in Experiment 1 (1a and 1c) and Experiment 2 (2a and 2c), or for the semantically-unrelated conditions in Experiment 1 (1b and 1d) and Experiment 2 (2b and 2d) (all p values > 0.25).Note.In Experiment 1, participants were instructed to "Look for insects, ignore fruits" (Conditions 1a, 1b) or "Look for tools, ignore fruits" (Conditions 1c, 1d) and there was no violation to the primary-task picture category since participants were presented with insects and fruits (Conditions 1a, 1b) or tools and fruits (Conditions 1c, 1d).In Experiment 2, participants were also instructed to "Look for insects, ignore fruits" but there was a violation to the primary-task picture category since the category of pictures was changed either to tools and fruits (Conditions 2a, 2b) or musical instruments and fruits (Conditions 2c, 2d).
of-fit test with an alpha level of 0.05 and a large effect size (w) of 0.64 indicated that a minimum of 13 participants was required in each condition to achieve a power of 0.90.Taking direction from an IB study where task expectations were violated (White & Aimola Davies, 2008; Experiment 1), the chi-square goodness-of-fit test with an alpha level of 0.05 and a large effect size (w) of 0.57 indicated that a minimum of 17 participants was required in each condition to achieve a power of 0.90.

Apparatus
Data was collected "in-lab" (Experiment 1: 50 %; Experiment 2: 10 %)-one-to-one with the experimenter in a laboratory room on the ANU campus, or remotely "on-zoom" (Experiment 1: 50 %; Experiment 2: 90 %)-using the video conferencing platform Zoom.For both collection methods, the experiment was conducted under the direction of the experimenter, who was either in the laboratory room with the participant, or on Zoom.The on-zoom participants were visible and audible to the experimenter on video for the duration of the experiment since they were required to share their screen, so that the experimenter had visibility of the participant's desktop display.During the experiment, participants could see only the trial display on their screen, but they could hear the instructions from the experimenter.
For in-lab participants, experiments were run on a Dell Intel® Core ™ i7-9700 computer and presented on a Dell Intel® UHD Graphics 630 monitor, with a frame rate of 59.94 Hz and a 2560 x 1440 px spatial resolution with a screen size of 53.42 cm.The experiment was programmed and presented in E-basic (E-Prime, V2, Psychology Software Tool Inc., USA).A chin-and-head rest fixed the viewing distance at 57 cm from the screen.
For on-zoom participants, a personal desktop or laptop computer was required, with sound, a microphone, a webcam, and internet connection.The average screen-size was 39.77 cm (range: 30.48-68.58 cm).The experiment was coded and presented using Inquisit Software V6.4.2.To approximately replicate in-lab conditions, the on-zoom participants were in a quiet space free from distractions, and they sat square-on-facing the monitor screen, positioned at an arms-length (approximately 57 cm) from the screen.The height of the screen was adjusted using books or boxes, ensuring the participant's eyes were in-line and level with the centre of the display.The viewing distance and position were adjusted under the strict supervision of the experimenter, as they were for the in-lab participants.

Stimuli
The stimuli for the experiments were identical for in-lab and on-zoom participation.All stimuli were designed using Inkscape Version 0.91.There were seven trials: four Non-Critical Trials, followed by a Critical Trial, a Divided-Attention Trial, and a Full-Attention Trial.The first four trials were Non-Critical Trials, and each contained three types of displays: (i) black fixation cross (each arm of the cross = .05• ) centred on a white circular disk (9.76 • in diameter) within a black background; (ii) primary-task display with four pictures, each positioned in a separate quadrant of a white circular disc with the far corners of each image 3.70 • from a centre cross; (iii) white screen.The Critical, Divided-Attention, and Full-Attention Trials contained the same sequence of three displays as the Non-Critical Trials, but in addition included a critical word (21-point Arial) in the centre of the white circular disc that appeared simultaneously with the four pictures in the primary-task display (see Fig. 1).The picture stimuli (average size: 2.1 • by 1.6 • ) were black line drawings of exemplars from four semantic categories: 12 insects, 12 fruits, 12 tools and 2 musical instruments.
The pictures from the semantic categories of insects, tools, fruits, and musical instruments were sourced from Snodgrass and Vanderwart (1980) and a free-to-use online database.To ensure a similar perceptual and cognitive load for each of the trials, with the aim of minimising variability across the experiment conditions of the two IB experiments, the pictures and critical words used in the two IB experiments were assessed in a preliminary study (see Supplementary Material for the complete report of this preliminary study, 'Category and Exemplar Norming Study').In this preliminary study, 110 neurologically healthy participants (80 females, M age = 20.66,SD age = 4.67, range age : 18-56) responded to a series of questions in an online Qualtrics survey evaluating the picture and critical word stimuli.The findings for the picture stimuli indicated there were high levels of category agreement, name agreement, familiarity and image agreement, and that the pictures were within the recommended middle-level visual complexity (see Chen et al., 2017).For example, the primary-task and distractor pictures chosen for the IB experiments had a category agreement rating of 80 % or higher, indicating strong consensus among participants regarding category exemplar membership (e.g., approximately 95 % agreement that a "butterfly" represents the category "insects").Three six-letter words-"BEETLE", "WRENCH" and "VIOLIN", exemplars of the categories of insects, tools and musical instruments, respectively-were chosen as the critical words (depending on the experiment condition).The comparable strength of the word associations (e.g., 0.07 for BEETLE, 0.08 for WRENCH, 0.13 for VIOLIN), and the proportion of participants generating each word (37 % for BEETLE, 31 % for WRENCH, 50 % for VIOLIN) in response to their corresponding category cues ('insects,' 'tools,' 'musical instruments'), suggests that each word was being spontaneously generated in association with its respective category cue.Furthermore, the chosen critical words had comparable frequency of use in the iWeb English Corpora and were aligned well on measures of semantic relatedness, such as Latent Semantic Analysis (Landauer & Dumais, 1997) and JCN (Jiang & Conrath, 1997), reinforcing their suitability for the experiment conditions.

Procedure
Each participant completed seven trials.Both in-lab and on-zoom participants received identical instructions from the experimenter, explaining the primary task before beginning the experiment.Two types of pictures were presented in the primary-task display of each trial: primary-task pictures and distractor pictures.Primary-task pictures were of key importance to the participants, since the task was to look for primary-task pictures, and to ignore distractor pictures.Depending on the condition, the participants' task was either to S. Chu and A. Aimola Davies "Look for insects, ignore fruits" or to "Look for tools, ignore fruits".For example, in one condition where the participants were tasked with "Look for insects, ignore fruits", they were given the following instructions verbatim:

First you will see a blank screen [blank slide displayed]. When I say "go" press the spacebar to start the trial. Please wait for me to say "go" because between trials I will be writing down your responses. After you hit the spacebar, a cross will appear [fixation slide displayed]. Fix your gaze on this cross. Then insects and fruits will appear briefly [pictures slide displayed]. Look for the insects, ignore the fruits. It is important for control purposes that neither of us speak when the pictures are on the screen. Wait for the pictures to go away and then when the blank screen appears [blank slide displayed], report aloud the names of the insects you just saw.
For all conditions, each of the seven trials was initiated by the experimenter saying "go", prompting the participant to press the spacebar.A fixation cross was presented for 1200 ms, followed by the primary-task display for a further 1200 ms, followed by a blank screen (see Fig. 2).Depending on the category that participants were instructed to look for, they were asked either "please report the insects" or "please report the tools".The experimenter manually recorded the participants' responses verbatim.A practice trial was conducted prior to the experiment trials, to ensure that the participants understood the task to look for and then report the insects (or Note.The sequence of displays for each trial included a fixation cross, a primary-task display, and a blank screen.To move from the blank screen to the beginning of the next trial, participants pressed the space bar (after having given a verbal response naming the primary-task pictures, either insects or tools).A critical word appeared at fixation simultaneously with the pictures in Critical Trial 5, Divided-Attention Trial 6, and Full-Attention Trial 7.
S. Chu and A. Aimola Davies tools).The successful report of the pictures in the practice trial confirmed that the participant understood the picture-naming task.
Critical Trial 5 was the critical trial for all participants.A critical word-BEETLE, WRENCH, or VIOLIN-appeared in the centre of the display, simultaneously with the primary-task pictures.In the critical trials, the assumption was that participants would be focused on the primary task of looking for and naming pictures (either insects or tools), and therefore their attentional resources would be directed to what they were told was the semantic category of the primary-task pictures.To determine whether a participant noticed the critical word, there were three questions probing detection, identification, and location with strict criteria applied to evaluate the participants' responses to each question.To determine detection, once participants verbally reported the primary-task pictures, they were asked: "Did you notice anything on this trial in addition to the insects (or tools) and fruits?".If participants responded "no", this was recorded and the participant moved on to the next trial in the experiment.To satisfy the detection criterion, participants were required to respond "yes", and were then asked two further questions, about identification and location.To determine identification, participants were asked, "What did you detect?" and then, to determine location, participants were asked "Where was it?".Participants' responses were recorded verbatim.To satisfy the identification criterion, participants were required to say the name of the word (e.g., BEETLE, WRENCH, or VIOLIN).To satisfy the location criterion, participants were required to respond verbally that the additional thing they detected was in the centre of the screen.Variations on the description of the centre were accepted, such as "middle" or "where the fixation cross used to be".Participants were coded as "noticers" or "non-noticers" of the critical word.To be coded as a noticer, participants were required to give a "yes" response to the question "Did you notice anything on this trial in addition to the insects (or tools) and fruits?", and also to provide either the correct identification (i.e., the critical word) or the correct location of the critical word.Of the 97 participants who responded "yes" to the question "Did you notice anything on this trial in addition to the insects (or tools) and fruits?", 95 participants were coded as "noticers": 85 of these participants met both the identification and the location criterion, six participants met the identification criterion but not the location criterion (Experiment Conditions 2b, 2c) and four participants met the location criterion but not the identification criterion (Experiment Conditions 1c, 1d, 2c).Two participants met neither criterion (Experiment Conditions 1b and 2a) and were coded as "non-noticers".The proportion of participants coded as noticers who met both the detection and the location criteron is 85/95 (89.5 %).
In Divided-Attention Trial 6, participants were presented with the same stimuli as in Critical Trial 5.In the divided-attention trials, the assumption was that participants were alerted to the fact that there would be something additional in the display (as this was alluded to with the question at the end of Critical Trial 5) and they would be dividing their attention between the primary task of looking for and naming pictures and the search for possible other information.Once participants verbally reported the primary-task pictures, they were asked the same three questions as in Critical Trial 5. Participants were coded as "noticers" or "non-noticers" of the critical word using the same criteria as in Critical Trial 5.
In Full-Attention Trial 7, participants were presented with the same stimuli as in the previous two trials, but they were given new instructions prior to the commencement of the trial: "On the next trial, do not worry about reporting the insects (or tools).Instead, tell me if you notice anything else on the screen."In the full-attention trials, the assumption was that participants would have full attentional resources to detect the critical word.Unlike the previous trials where attention was mainly on the primary task of looking for and naming pictures, in Full-Attention Trial 7, participants were explicitly told to pay attention to anything else in the display.Participants were coded as "noticers" or "non-noticers" of the critical word using the same criteria as in Critical Trial 5 and Divided-Attention Trial 6.
After the IB experiment, participants completed the Edinburgh Handedness Inventory and a Demographic and Language History Questionnaire, either in hard copy for participants in-lab or in Qualtrics for participants on-zoom.Participants were then asked about any prior experience and knowledge of IB.Finally, the participants were debriefed and compensated for their time.

Experiment design
Experiment 1 (four experiment conditions) was a between-subjects design: as in all IB experiments, each participant can only experience Critical Trial 5 once, thus providing a single data point.The dependent variable was the proportion of noticers for the critical word.
We aimed to investigate whether a critical word that was semantically related to the primary-task picture category would be noticed more than a semantically-unrelated critical word.In the first two conditions of Experiment 1, participants were instructed to "Look for insects, ignore fruits".In Critical Trial 5, Divided-Attention Trial 6, and Full-Attention Trial 7, participants received either the semantically-related critical word BEETLE (Condition 1a: Primary-Task Picture Category: Insects, No-Violation-Semantically-Related), or the semantically-unrelated critical word WRENCH (Condition 1b: Primary-Task Picture Category: Insects, No-Violation-Semantically-Unrelated; see Fig. 1).In the other two conditions, participants were instructed to "Look for tools, ignore fruits".In Critical Trial 5, Divided-Attention Trial 6, and Full-Attention Trial 7, participants received either the semantically-related critical word WRENCH (Condition 1c: Primary-Task Picture Category: Tools, No-Violation-Semantically-Related), or the semantically-unrelated critical word BEETLE (Condition 1d: Primary-Task Picture Category: Tools, No-Violation-Semantically-Unrelated; see Fig. 1).Participants were randomly assigned to one of these four Conditions: 1a, 1b, 1c, or 1d (see Table 1).

Statistical analysis plan
All statistical analyses were conducted using IBM SPSS Statistics Version 29.0 for Windows with an alpha level of 0.05.

Primary-task accuracy for noticers and non-noticers
Two-tailed Mann-Whitney U Tests were used to investigate primary-task accuracy.Primary-task accuracy was measured by summing the correctly identified primary-task pictures across Non-Critical Trials 1-4.A difference between noticers and non-noticers of the critical word in their performance of the primary task might suggest an alternative explanation of why some participants failed to report the critical word.For example, if noticers performed the primary task more accurately than non-noticers, this might suggest that non-noticers did not fully engage with the experiment.On the other hand, if non-noticers performed the primary task more accurately than noticers, this might suggest that non-noticers attended to the primary task more effortfully than noticers and consequently did not have spare cognitive resources available to process the critical word.

Semantic relatedness between the critical word and the primary-task picture category
Inattentional blindness was analysed using log-linear analysis and chi-square tests.Assumptions of raw frequency data, random sampling, and independence of observations were all met.Additionally, the assumption that all expected cell frequencies were greater than five was met for the analyses of Critical Trial 5 and Divided-Attention Trial 6 (Tabachnick & Fidell, 2019).When expected cell frequencies were less than five in Full-Attention Trial 7, the Yates continuity correction was used (Yates, 1934).The likelihood ratio chi-square statistic, G 2 , was reported for the log-linear analysis.For all chi-square analyses, χ 2 , the Pearson chi-square statistic (Howell, 2010), with two-tailed p-values, odds ratios with 95 % confidence intervals were reported.Note.All trials included a display of 4 pictures.Non-Critical Trials 1-4 included either 1 primary-task picture and 3 distractor pictures or 3 primarytask pictures and 1 distractor picture.Critical Trial 5, Divided-Attention Trial 6, and Full-Attention Trial 7 included 2 primary-task pictures and 2 distractor pictures.Conditions 1a and 1b had the same primary-task-picture and distractor-picture categories: insects and fruits.Conditions 1c and 1d had the same primary-task-picture and distractor-picture categories: tools and fruits.
A three-way loglinear analysis-2 'Primary-Task Picture Category' (insects, tools) x 2 'Critical Word' (BEETLE, WRENCH) x 2 'Noticers' (noticers, non-noticers)-was conducted to compare the proportion of noticers for the critical word in Critical Trial 5.The significant three-way interactions in the loglinear analyses were followed up with two-way chi-square tests.Instead of collapsing the data by semantic relatedness (i.e., semantically-related and semantically-unrelated critical words), separate analyses were conducted to rule out the possibility of potential influences due to the primary-task picture category.Two separate two-way chi-square tests were conducted for the two different task instructions "Look for insects, ignore fruits" and "Look for tools, ignore fruits".All significant twoway interactions and main effects were reported (p < 0.05).These analyses were conducted separately for Critical Trial 5, Divided-Attention Trial 6, and Full-Attention Trial 7.

In-lab and on-zoom data collection methods
Due to a state-wide lockdown in response to the COVID-19 pandemic in August 2021, in-person data collection was halted, and the remainder of participants in the present study were tested on the video conferencing platform Zoom.This involved running the paradigm on the participants' personal computers from their home, under the direction of the experimenter.For each of the four conditions in Experiment 1, 50 % of the data was collected in-lab and 50 % was collected on-zoom.As Experiment 1 had large numbers of participants for both in-lab and on-zoom data collection methods, this provided the opportunity to test whether the collection method had an effect on the proportion of noticers for the critical word; and importantly, whether IB was a robust effect, replicable outside of strict lab conditions.Therefore, a 2 'Collection Method' (in-lab, on-zoom) x 2 'Noticers' (noticers, non-noticers) chi-square test was used to analyse the proportion of noticers for the critical word in Critical Trial 5 for Conditions 1a, 1b, 1c, and 1d.

Semantic relatedness between the critical word and the primary-task picture category
For Critical Trial 5 of Experiment 1, a three-way hierarchical loglinear analysis comparing the proportion of noticers across all four conditions produced a final model that retained all effects.This indicated that the three-way interaction, 'Critical Word' (BEETLE, WRENCH) x 'Primary-Task Picture Category' (insects, tools) x 'Noticers' (noticers, non-noticers) was significant, G 2 (1, N = 96) = 37.64, p < 0.001.The two-way chi-square test results for Critical Trial 5 (see Table 2, comparisons for Critical Trial 5: Conditions 1a and 1b, and Conditions 1c and 1d) indicated that when participants were instructed to "Look for insects, ignore fruits", there was a significantly higher proportion of noticers for the semantically-related critical word BEETLE (noticers = 83.3%) than for the semantically-unrelated critical word WRENCH (noticers = 20.8%).The odds of noticing the critical word were 19.00 times higher for BEETLE, semantically related to the primary-task picture category of insects, than for the semantically-unrelated critical word WRENCH.Similarly, when participants were instructed to "Look for tools, ignore fruits", there was a significantly higher proportion of noticers for the semantically-related critical word WRENCH (noticers = 79.2%) than for the semantically-unrelated critical word BEETLE (noticers = 20.8%).The odds of noticing the critical word were 14.44 times higher for WRENCH, semantically related to the primary-task picture category of tools, than for the semantically-unrelated critical word BEETLE.Fig. 3 depicts the significantly higher proportion of noticers for the critical word semantically related to the primary-task picture category as compared to the semantically-unrelated critical word, for both attended categories-insects and tools in Critical Trial 5.
For Divided-Attention Trial 6 of Experiment 1, a three-way hierarchical loglinear analysis comparing the proportion of noticers across all four conditions produced a final model that retained all effects.This indicated that the three-way interaction, 'Critical Word' (BEETLE, WRENCH) x 'Primary-Task Picture Category' (insects, tools) x 'Noticers' (noticers, non-noticers) was significant, G 2 (1, N = 96) = 10.89,p < 0.001.The two-way chi-square test results for Divided-Attention Trial 6 comparisons-Conditions 1a and 1b ("Look for insects, ignore fruits") and Conditions 1c and 1d ("Look for tools, ignore fruits")-indicated the same pattern of results as for Critical Trial 5 (see Table 2).There was a significantly higher proportion of noticers for a critical word semantically related to the primary-task picture category that the participants were instructed to "Look for" compared to a semantically-unrelated critical word (see Fig. 3).
For Full-Attention Trial 7 of Experiment 1, a three-way hierarchical loglinear analysis comparing the proportion of noticers across all four conditions produced no significant three-way or two-way interaction effects but there was a significant main effect of 'Noticers' G 2 (1, N = 96) = 60.74,p < 0.001.A one-way chi-square test across all conditions indicated a significant difference between participants who reported the critical word (noticers = 87.5 %) and participants who did not report the critical word (non-noticers = 12.5 %), χ 2 (1, N = 96) = 54.00,p < 0.001.No other main effects or two-way interactions for Experiment 1 reached significance (all ps > .09).
Overall, the analyses in Experiment 1 indicated that participants were more likely to notice a critical word semantically-related to the primary-task picture category compared to a semantically-unrelated critical word.This pattern was consistent across Critical Trial Note. a Each odds ratio compared the odds of noticing the critical word in a pair of experiments.b An example of how to interpret the Odds Ratio and 95% Confidence Interval columns is as follows: In Critical Trial 5, the comparison between Condition 1a and Condition 1b, where participants were instructed to "Look for insects, ignore fruits", the odds of noticing the critical word were 19 times higher when it was semantically related to the primary-task picture category than when it was semantically unrelated (WRENCH; 95% CI [4.42-81.57]).*p < .05. **p < .01. ***p < .001.

S. Chu and A. Aimola Davies
5 and Divided-Attention Trial 6, with an increase in noticers in Divided-Attention Trial 6 compared to Critical Trial 5.For Full-Attention Trial 7, there were no significant interaction effects, but there was a significant difference between participants who noticed the critical word and those who did not.

In-lab and on-zoom data collection methods
The results of the 2 'Collection Method' (in-lab, on-zoom) x 2 'Noticers' (noticers, non-noticers) chi-square test demonstrated no significant difference for the proportion of noticers across Conditions 1a, 1b, 1c, and 1d for participation in-lab compared with participation on-zoom, χ 2 (1, N = 96) = 1.04, p = 0.31, OR = 1.52 (95 % CI: 0.68, 3.40).It was assumed that there would continue to be no significant difference between the collection methods for the remainder of the study, and participant data for Experiment 2 continued to be collected on Zoom.

Experiment design
Experiment 2 (four experiment conditions) was also a between-subjects design, and the dependent variable was the proportion of noticers for the critical word.
The aims of Experiment 2 were similar to the aims of Experiment 1, in that we aimed to to investigate whether a semanticallyrelated critical word would be noticed more than a semantically-unrelated critical word.The main difference between Experiments 1 and 2 was that in Experiment 2 we also aimed to investigate whether a critical word that was semantically related to an unexpected picture category would be noticed more than a semantically-unrelated critical word.For example, in Experiment 2, although all participants were instructed to "Look for insects, ignore fruits", the primary-task pictures of insects were unexpectedly replaced with pictures of tools (or musical instruments).We investigated whether, for example, the critical word WRENCH, which was semanticallyrelated to the unexpected pictures of "tools" would be noticed more than the semantically-unrelated critical word VIOLIN.
All participants in Experiment 2 were instructed to "Look for insects, ignore fruits".In Critical Trial 5, Divided-Attention Trial 6, and Full-Attention Trial 7, all participants experienced a violation to their task expectations for the picture category in the primary task.In the first two conditions, the primary-task picture category was unexpectedly changed from insects to tools.Participants received either the semantically-related critical word WRENCH (Condition 2a: Primary-Task Picture Category: Insects, Picture Violation with Tools-Semantically-Related), or the semantically-unrelated critical word VIOLIN (Condition 2b: Primary-Task Picture Category: Insects, Picture Violation with Tools-Semantically-Unrelated; see Fig. 1).In the other two conditions, the primary-task picture category was unexpectedly changed from insects to musical instruments.Participants received either the semantically-related critical word VIOLIN (Condition 2c: Primary-Task Picture Category: Insects, Picture Violation with Musical Instruments-Semantically-Related) or the semantically-unrelated critical word WRENCH (Condition 2d: Primary-Task Picture Category: Insects, Picture Violation with Musical Instruments-Semantically-Unrelated; see Fig. 1).Participants were randomly assigned to one of these four Conditions: 2a, 2b, 2c, or 2d (see Table 3).

Statistical analysis plan
All statistical analyses were conducted using IBM SPSS Statistics Version 29.0 for Windows with an alpha level of 0.05.

Primary-task accuracy for noticers and non-noticers
Two-tailed Mann-Whitney U Tests were used to investigate primary-task accuracy (see Experiment 1, for a full explanation).

Violation of expectations for the primary-task picture category
Inattentional blindness was analysed using log-linear analysis and chi-square tests (see Experiment 1, for a full explanation of assumptions, random sampling, and independence of observations).The likelihood ratio chi-square statistic, G 2 , was reported for the log-linear analysis.For all chi-square analyses, χ 2 , the Pearson chi-square statistic (Howell, 2010), with two-tailed p-values, odds ratios with 95 % confidence intervals were reported.
A three-way loglinear analysis-2 'Primary-Task Picture Violation' (tools, musical instruments) x 2 'Critical Word' (WRENCH, VIOLIN) x 2 'Noticers' (noticers, non-noticers)-was conducted to compare the proportion of noticers for the critical word in Critical Trial 5.The significant three-way interactions in the loglinear analyses were followed up with two-way chi-square tests.Instead of collapsing the data by semantic relatedness (i.e., semantically-related and semantically-unrelated critical words), separate analyses were conducted to rule out the possibility of potential influences due to the unexpected picture category.Two separate two-way chi-square tests were conducted for the two different unexpected picture categories "tools" and "musical instruments".All significant two-way interactions and main effects were reported (p < 0.05).These analyses were conducted separately for Critical Trial 5, Divided-Attention Trial 6, and Full-Attention Trial 7.

Violation of expectations for the primary-task picture category
For Critical Trial 5 of Experiment 2, a three-way hierarchical loglinear analysis comparing the proportion of noticers across all four Note.a Each odds ratio compared the odds of noticing the critical word in a pair of experiments.b An example of how to interpret the Odds Ratio and 95% Confidence Interval columns is as follows: In Critical Trial 5, the comparison between Condition 2a and Condition 2b, where participants were instructed to "Look for insects, ignore fruits", the odds of noticing the critical word were 5 times higher when it was semantically related to the unexpected picture category than when it was semantically unrelated (WRENCH;).*p < .05. **p < .01. ***p < .001.

Violation or no violation of expectations for the primary-task picture category-a comparison
The results of Experiment 2 demonstrated that, when participants' expectations for the primary-task picture category were violated, there was a higher proportion of noticers for the critical word semantically-related to the new and unexpected picture category compared to a semantically-unrelated critical word.This suggests that a violation to participants' expectations for the primary-task pictures alone was not sufficient to reorient attention to the critical word.
An additional analysis comparing two experiment conditions was conducted to test whether there was a difference in the proportion of noticers for the critical word WRENCH when participants' expectations for the primary-task picture category were violated (Experiment 2d) compared to when their expectations were met (Experiment 1b).In both experiment conditions, the critical word was not semantically related to either of the picture categories presented to participants.In Experiment 1b, participants were asked to "Look for insects, ignore fruits" and the pictures presented in Critical Trial 5 were as they expected-participants were presented with insects and fruits.The findings indicated that only 20.8 % of participants reported noticing the critical word when their expectations for the primary-task picture category were met.In Experiment 2d, participants were asked to "Look for insects, ignore fruits" but the pictures presented in Critical Trial 5 were not as they expected-participants were presented with musical instruments and fruits.The findings indicated that 33.3 % of participants reported noticing the critical word when their expectations for the primary-task picture category were violated.Therefore, this additional comparison was included as a direct test of whether there was a difference in the proportion of noticers for a semantically-unrelated critical word when participants' expectations for the primary-task picture category were met (Experiment 1b) as compared to when the participants' expectations for the primary-task picture category were violated (Experiment 2d).
A two-way chi-square test-2 'Violation Type' (no violation, violation with musical instruments) x 2 'Noticers' (noticers, non-noticers)was conducted for Critical Trial 5, to compare the proportion of noticers for a critical word in Experiment 1b (No-Violation-Semantically-Unrelated) with a critical word in Experiment 2d (Violation with Musical Instruments-Semantically-Unrelated).All participants were presented with the critical word WRENCH, which was semantically unrelated to the task instructions "Look for insects, ignore fruits", and was unrelated to the expected pictures of insects presented in Experiment 1b or the unexpected pictures of musical instruments presented in Experiment 2d.The chi-square test indicated that there was no significant difference between the proportion of noticers for the critical word WRENCH in Critical Trial 5 for Experiment 1b-no violation of participants' expectations for the picture category "insects" (noticers = 20.8%) and Experiment 2d-violation of participants' expectations for the picture category "insects" with the picture category changed from insects to musical instruments (noticers = 33.3%), χ 2 (1, N = 48) = 0.095, p = 0.33, OR = 0.53 (95 % CI: 0.14, 1.93). 3

Discussion
In the present study, we used a static IB paradigm, similar to Koivisto and Revonsuo (2007) to investigate the effects of semantic relatedness and expectation on attention.In Experiment 1, we used two new, not previously investigated primary-task picture categories-insects and tools.We found that although all participants were presented with an unexpected six-letter critical word (21-point Arial) in the centre of the screen in Critical Trial 5, a higher proportion of participants reported a critical word semantically related to the primary-task picture category than a semantically-unrelated critical word.Thus, we replicated the 'semantic congruency effect' previously found by Koivisto andRevonsuo (2007, 2009).This finding was consistent across both of the primary-task picture categories-insects and tools.For example, when participants were instructed to "Look for insects, ignore fruits", a significantly higher proportion of participants reported the critical word BEETLE compared to WRENCH, and when instructed to "Look for tools, ignore fruits", a significantly higher proportion of participants reported the critical word WRENCH compared to BEETLE.The results of Experiment 1 highlight the role of top-down attentional control in guiding participants' attention toward the instructed semantic category.Top-down attentional processes are considered to be influenced by expectation or voluntary control, so that attention can be directed toward task goals (Corbetta et al., 2000;Corbetta & Shulman, 2002).
In Experiment 2, we directly violated participants' expectations for the primary-task picture category.In Critical Trial 5, an unexpected picture category was presented to participants instead of the category they were instructed to look for.We found that although all participants were given the instructions "Look for insects, ignore fruits" and were presented with an unexpected picture category, a higher proportion of participants reported a critical word semantically related to the unexpected picture category than a 3 Some theoretical interest attaches to the finding that there was no significant difference between the proportion of noticers for the critical word when there was no violation of participants' expectations for the picture category and the proportion when there was a violation.The finding is consistent with the claim that a violation of participants' expectations for the picture category does not in and of itself affect whether participants report the critical stimulus.To investigate whether the finding supported the null hypothesis, we carried out additional Bayesian contingency table tests (an equivalent to the frequentist chi-square tests) to examine the relative strength of the evidence for the null hypothesis (H 0 ) and alternative hypothesis (H 1 ).These Bayesian statistical analyses were conducted using JASP Version 0.17.1 for Mac, employing an independent multinomial sampling and a default prior concentration of 1.Our H 0 was that there would be no difference between the proportion of noticers for the critical word WRENCH when there was no violation of participants' expectations for the picture category ("insects" in Experiment 1b) and when there was a violation of participants' expectations for the picture category ("musical instruments" in Experiment 2d), while our H 1 was that there would be such a difference.The results revealed a Bayes Factor favouring H 0 of BF 01 = 2.07 and a median log odds ratio = − 0.61, 95% credible interval [-1.87, 0.65].These findings indicated that the data was 2.07 times more likely to support H 0 compared to H 1 , providing evidence in favour of there being no difference in the proportion of noticers for the critical word when there was no violation of participants' expectations for the picture category and the proportion when there was a violation.We are grateful to a reviewer for suggesting this additional analysis.critical word that was unrelated to the picture category.This finding was consistent across both of the unexpected picture categories-tools and musical instruments.For example, when participants were instructed to "Look for insects, ignore fruits" and the unexpected picture category was tools, a significantly higher proportion of participants reported the critical word WRENCH compared to VIOLIN.Similarly, when participants were instructed to "Look for insects, ignore fruits" and the unexpected picture category was musical instruments, a significantly higher proportion of participants reported the critical word VIOLIN compared to WRENCH.The results of Experiment 2 demonstrate that violation of task expectations does not, in and of itself, lead to reorienting of attention toward the critical word presented at fixation.Instead, it was the semantic relatedness between the unexpected picture category and the critical word that led to reorienting of attention to the critical word.
In Experiment 1 of the present study, as in the seminal study of Koivisto and Revonsuo (2007), there was a significant semanticcongruency effect.Participants were more likely to report the critical stimulus-an unexpected word at fixation-if it was semantically related to the primary-task picture category than if it was unrelated.Koivisto and Revonsuo (2009) proposed an explanation for this effect, beginning from the hypothesis of "automatic processing of the meaning [of the critical word]" (p.865).On this account, the critical word-whether semantically related or unrelated to the primary-task picture category-would be automatically processed at least up to the level of semantic category.If the critical word was semantically related to the picture category then it would already be prioritised for attention as a result of top-down guidance of attention toward the instructed category (e.g., "Look for insects").Consequently, the critical word would be likely to be reported.If the critical word was unrelated to the picture category then it would not be prioritised for attention and would be less likely to be reported.
Sometimes, a critical word unrelated to the primary-task category and not prioritised for attention is reported.Here we appeal to Lavie's (1995Lavie's ( , 2005;;Lavie et al., 2014) perceptual load theory.The limited capacity of perception may be fully occupied by primarytask processing but, if perceptual load is low, then processing capacity may "spill over" (see Lavie, 2005, p. 75) to items that are not related to the primary task-such as an unexpected word that is not already prioritised for attention.
Awareness will thus be clearly confined to just the attended task information under conditions of high perceptual load (allowing for early selection effects of top-down attentional selection).In conditions of low load, however, awareness will not be confined to just those stimuli specified by top-down selection settings as 'to be attended to'.(Lavie et al., 2014, p. 2) Koivisto and Revonsuo (2009) manipulated perceptual load by presenting six pictures in the high load condition and only two pictures in the low load condition; in both conditions, only one picture (of a dog or a table) needed to be named (see p. 866).They found-consistent with perceptual load theory-that nearly a third of participants detected the semantically-unrelated critical word in the low load condition, but only 7 % in the high load condition: "increasing perceptual load almost completely eliminated the detection of semantically [unrelated] stimuli" (p.867).It is plausible that in our Experiment 1, with four pictures presented, perceptual load was not as high as in Koivisto and Revonsuo's (2009) high load condition and, in fact, 21 % of participants reported the semanticallyunrelated critical word.

Violation of expectations
The explanation of the findings in Experiment 1 of the present study appeals to two factors: (1) top-down guidance of attention toward the instructed category; and (2) perceptual load that was sufficiently low to allow some processing of stimuli that were not already prioritised for attention.Can the explanation be extended to Experiment 2, in which all four conditions involved expectation violation?It is important that, in Experiment 2, the critical word (VIOLIN or WRENCH) was always unrelated to the instructed primary-task picture category, which was insects in all four conditions ("Look for insects, ignore fruits").Participants were significantly more likely to report the critical word (e.g., VIOLIN) if it named an exemplar of the unexpected picture category (e.g., musical instruments) rather than an exemplar of a different category (e.g., tools).This difference in reporting rates cannot be explained by appeal to factor (1), top-down guidance of attention toward the instructed category.
We propose that the unexpected picture category is prioritised for attention as a result of the expectation violation.An inchoate sense that 'something unexpected is happening' may result in a generalised increase in vigilance-which might make it more likely that a critical word would be reported independently of its semantic relatedness to the picture category.The expectation violation in Experiment 2, in contrast, has a clear locus.The top-down expectation that pictures of insects will be presented is violated, bottom-up, by the absence of pictures of insects and the presentation of pictures of musical instruments (or pictures of tools) instead.Attention is directed to this locus of expectation violation and, specifically, to the unexpected picture category.A critical word semantically related to the unexpected picture category is also prioritised for attention as a result of the expectation violation, whereas a semantically unrelated critical word is not prioritised for attention.As in Experiment 1, the finding that a semantically unrelated critical word was sometimes reported in Experiment 2 can be explained by appeal to factor (2), perceptual load that was sufficiently low to allow some processing of stimuli that were not already prioritised for attention.We can unify the explanations of Experiment 1 and Experiment 2 by generalising factor (1) to encompass both (a) top-down guidance of attention toward the instructed category and (b) guidance of attention toward the unexpected category driven by expectation violation.
Expectation violations in IB have also been investigated by White and Aimola Davies (2008) and it may appear that the consequences of expectation violations in the White and Aimola Davies study contrast with those in the present study.White and Aimola Davies found that expectation violations reduced IB for the critical stimulus whereas, in the present study, expectation violations reduced IB only when the critical word was semantically related to the unexpected picture category.Semantically unrelated critical words remained largely unreported.
Before considering the apparent contrast in findings, we give a brief description of the IB paradigm used by White and Aimola Davies (2008).The primary-task objects to be named were always one, two or three letters that were presented on the arms of a cross and there were no distractors-that is, no to-be-ignored objects.The critical stimulus was always a small square, similar in size to the letters, and it appeared in one of the four quadrants of the cross.The display (cross and letters, and square on the critical trial) was presented for 200 ms.Attention was guided top-down to the category of letters ("Name the letters") but the critical square was not semantically related to the category of letters and was not prioritised for attention (though see further below).The expectations to be violated concerned the number of letters that would be presented on the arms of the cross and those expectations were generated by a numerical cue that was shown before the display.The cue was either valid (i.e., the number of letters presented on the arms of the cross was as indicated by the cue) or else invalid (i.e., the number of letters presented was one fewer-over-cueing or one more-undercueing than indicated by the cue).
The key finding of the White and Aimola Davies (2008) study was that, when numerical expectations were violated (invalid cue), IB was significantly reduced-more participants reported the critical square than when numerical expectations were met (valid cue). 4 The authors' explanation for this finding began from the hypothesis that "numerical expectation serves as a cue to termination" (p.1094).If a cue was valid then participants "may have terminated processing once the relevant number of items had been identified, thus preventing detection of further irrelevant items in the scene" (p.1101).Further processing of the display was unnecessary because there were no other stimuli prioritised for attention.If the cue was invalid then participants "may have attended to the array beyond the requirements of the primary task" (ibid.).It is of interest that, when no numerical cue was presented and no numerical expectation was established (Experiment 6), the rate of reporting the critical stimulus was high-not significantly different from the rates of reporting when a numerical cue was invalid.Thus, when the number of primary-task letters was uncertain (because there was no numerical cue, or an invalid over-cue or under-cue), participants continued processing the display and were likely to report the critical square.
If continued processing of the display resulted in substantially increased reporting of the critical square then-we assume-the perceptual load of the primary letter-naming task was sufficiently low to allow some processing of stimuli that were not already prioritised for attention.This assumption is plausible as there were only two (or three) letters on the arms of the cross and no distractors.This low level of perceptual load would also allow processing capacity to "spill over" to the critical stimulus in valid-cue conditions-and that was what White and Aimola Davies (2008) found.The difference in reporting rates between valid-and invalid-cue conditions was not explained by a difference in perceptual load, nor by a difference in semantic relatedness, but by the difference between terminated and continued processing of the display.
An alternative explanation for the high proportion of participants reporting the critical stimulus when numerical expectations were violated brings us closer to Koivisto andRevonsuo (2007, 2009) and Experiment 1 in the present study.There was a degree of uncertainty about the number of primary-task letters and participants had the attentional resources available to process the full visual display.Spare attentional capacity spilled over to the critical stimulus, a square that was visually similar to a letter "D" in both size and shape-especially given the very brief 200 ms presentation of the display.Consequently, the square was incorporated into the semantic category of letters and processed as if it were semantically related to the primary-task letters-even if explicit report would identify it as a square.

Semantic relatedness and the visual world paradigm
Research on IB that involves attention to semantic categories, including the present study, is informed by a body of research using the visual world paradigm (see Huettig et al., 2011 for a review) to investigate the influence of semantic relationships on participants' direction of visual attention.For example, in a study by Huettig and Altmann (2005), participants heard a target word (e.g., DESK) while they viewed a display comprising four pictures of objects.In the 'target' condition, one object (e.g., a desk) corresponded to the target word and participants were more likely to look at the target object than at unrelated distractors.In the 'competitor' condition, the target object was replaced by a competitor from the same semantic category (e.g., a bed) and participants were more likely to look at the within-category competitor object than at the distractors.In the 'target & competitor' condition, both the target object and the competitor object appeared in the display and participants were more likely to look at the target object than at the competitor object and more likely to look at the competitor object than at the distractors.Thus, semantic relatedness (membership in the same semantic category) influenced the direction of visual attention.
The findings of Yee and Sedivy (2006, Experiment 1) were similar to those in Huettig and Altmann's (2005) 'target & competitor' condition.That is, after hearing the target word (e.g., PIANO), participants were most likely to look at the target object (e.g., a piano), but also more likely to look at an object (e.g., a trumpet) semantically related to the target word than at an unrelated control object (e. g., a lamp).In the Yee and Sedivy study, the pairs of objects included, not only taxonomically-related pairs (from the same semantic category), but also thematically-related pairs (e.g., grapes-wine; hammer-nail).Thus, after hearing the word HAMMER, participants were more likely to look at a nail than at an unrelated couch.
The visual world display used by Huettig and McQueen (2007, Experiment 1) comprised pictures of a phonological competitor (same first two phonemes), a shape competitor (similar visual shape), and a semantic competitor (related in meaning) of the target 4 When only one letter was presented (low perceptual load conditions), more than half the participants reported the square when the numerical cue was valid and the difference in reporting rates between valid-and invalid-cue conditions was not significant.Also, when two letters were presented in an invalid-cue condition and the display duration was reduced to 125 ms, participants were still able to name the letters but less than half reported the square.object, along with an unrelated distractor object.(The target object did not appear in the display.)The target word was presented in a context sentence and was preceded by about seven words.The display appeared as the context sentence began.About 240 ms after the acoustic onset of the target word, the probability of fixation on the phonological competitor began to diverge from the probability of fixation on the distractor; fixations on the shape and semantic competitors diverged from fixations on the distractor about 100 ms later.Thus, during the 300-399 ms time window, there were significantly more fixations on the phonological competitor than on the distractor.During the next two time windows (400-599 ms), all three competitor objects attracted more fixations than the distractor but from 600 ms on, only the shape and semantic competitors were looked at more than the distractor.In short, overt attentional shifts to the phonological competitor were earlier than those to the semantic and shape competitors, consistent with spoken word recognition involving cascaded processing in which information about the sound of a word leads to information about what the word means and what the corresponding object looks like.No significant differences in time-course were found between looking at the semantic competitor and at the shape competitor.
As the authors noted, the cascaded processing in picture recognition proceeds from information about visual features to information about what object is depicted and what its name is.Thus, it was a crucial feature of Huettig and McQueen's (2007) Experiment 1 that there was sufficient time for the names of the pictures to be accessed before the target word was heard.In their Experiment 2, the display appeared only 200 ms before the acoustic onset of the target word and there was no difference between fixations on the phonological competitor and on the distractor.Chow et al. (2017) used the four-picture visual world paradigm to investigate in toddlers the time-course of overt attention to pictures phonologically or semantically related to a spoken word.The target word (e.g., BEE) was presented (in the context "Look at the_____!") 4000 ms after onset of the display, which comprised pictures of a phonological competitor (e.g., a bus) and a semantic competitor (e.g., a cat), along with two unrelated distractors (e.g., a sandwich and a dress).In Experiment 1, the semantic competitor was, specifically, taxonomically related to the target word and the participants were 24-and 30-month-olds (two groups).For both groups, fixations on the phonological competitor peaked earlier than fixations on the semantic (taxonomic) competitor.Fixations on the phonological competitor were greater than fixations on unrelated distractors across the interval 330-1130 ms after target word onset, whereas fixations on the taxonomic competitor were greater than fixations on the distractors from 630 ms onwards.In Experiment 2, the semantic competitor was thematically (and not taxonomically) related to the target word and participants were between 25-and 30-months-old.As in Experiment 1, fixations on the phonological competitor peaked earlier than fixations on the semantic competitor but participants' fixations on the thematic competitor in Experiment 2 increased more steeply and reached a higher peak than fixations on the taxonomic competitor in Experiment 1. Together, the two experiments showed that, following a spoken target word, toddlers' attention can be directed to phonologically-related and semantically-related pictures in a four-picture display in less than 1000 ms.The findings are strikingly similar to those for adults (Huettig & McQueen, 2007), now extended to a much lower age (24-30 months).
This body of research (Chow et al., 2017;Huettig & Altmann, 2005;Huettig & McQueen, 2007;Yee & Sedivy, 2006) shows that semantic-taxonomic or thematic-relationships between objects influence participants' direction of visual attention.Informed by this work, several IB studies, including Koivisto and Revonsuo (2007) and the present study, have shown that a critical stimulus taxonomically related to the primary-task objects is more likely to be reported than an unrelated critical stimulus.As far as we know, no IB study has demonstrated a similar effect for thematic relatedness.

Online inattentional blindness research
A noteworthy but unplanned contribution of the present study is the direct comparison of in-lab and on-zoom data collection methods in Experiment 1. Due to a state-wide lockdown in response to the COVID-19 pandemic in Canberra, in-person data collection was halted and instead continued online on Zoom, one-to-one with the experimenter.The comparison of the findings for the four experiment conditions tested with these two methods of data collection indicates there was no significant difference in the proportion of noticers for the critical word, whether participants were tested in-lab or on-zoom.Although the procedure and set-up for on-zoom participants replicated those used for in-lab participants, the environment for on-zoom participants was different each time as the size of the computer screen, size of the room, décor, and other features of the participants' home environments were unable to be controlled.As such, the consistent results across the two data-collection methods demonstrate the ecological validity of the findings in the present study.To the best of our knowledge, our study is the first to provide a comparison between in-lab data collection conducted in-person versus on-zoom data collection conducted online under the direction of an online experimenter, and thus adds to the literature demonstrating the reliability of data collected using online methods to investigate IB.In previous online IB data collection, participants independently completed the experiment without the experimenter (Kreitz et al., 2015;Redlich et al., 2021b;Stothart et al., 2015;Ward & Scholl, 2015;Wood & Simons, 2017;2019).As the IB paradigm used in the present study has not yet been tested with an online method, in which the task is completed without the involvement of the experimenter, further investigation is needed to assess its suitability.

Future directions and limitations
The present study provides avenues for future research to explore the effects of expectation and the violation of expectation on attention, and the effect of other contextual relationships on attention.Our first two suggestions for future research can be investigated with a behavioural study and the third with a neuroimaging study.
Future research could use a behavioural study to investigate systematically how an expectation violation reduces IB for a critical stimulus.Specifically, we are interested in whether the differences in findings between the study by White and Aimola Davies (2008) and the present study-in conditions where participants' expectations were violated-may be due not only to differences between the two paradigms, but also to the nature of the expectation violation.Future work could investigate whether the semantic-congruency effect remains when numerical expectations are violated.For example, researchers could set a numerical expectation for the picture-naming task in the present study by cueing the number of insects (or tools) to name in each trial.The expectation for the number of pictures-to-be-named could then be violated by displaying a number of pictures incongruent with the preceding cue.The researchers could thus compare noticers for critical words semantically-related and semantically-unrelated to the primary-task picture category under different numerical violation conditions.If the selection of stimuli for attention is driven by numerical expectations, we would predict outcomes similar to those observed by White and Aimola Davies: a higher proportion of noticers for the critical word when numerical expectations are violated compared to when numerical expectations are met (regardless of whether the critical word is semantically related to the category of primary-task pictures).In contrast, if the selection of stimuli for attention is driven by semanticcategory expectations, we would predict outcomes similar to those observed in the present study: a higher proportion of noticers for the critical word semantically related as compared to semantically unrelated to the picture category (regardless of whether numerical expectations are violated).
A second avenue for future research could use a behavioural study to investigate IB for stimuli with other contextual relationships.Specifically, we are interested in whether attention reorients to an unexpected phonologically-related critical word in an IB paradigm.Previous research using the visual world paradigm (Huettig & McQueen, 2007) found early attentional shifts to pictures and words phonologically related to a 'target' word spoken in a sentence.Future work could investigate IB for a critical word phonologically related to the participants' primary task.Using a similar experiment set-up as in Experiment 1 of the present study, the primary-task pictures to name in Critical Trial 5 would include exemplar pictures where the names of the pictures have the same initial phonemes as a phonologically-related critical word.The researchers could thus compare noticers for a phonologically-related critical word, a semantically-related critical word, and an unrelated critical word.If the selection of stimuli for attention is driven by phonological relationships, we would predict a higher proportion of noticers for the phonologically-related critical word compared to the semantically-related or unrelated critical word.In contrast, if the selection of stimuli for attention is driven by semantic-category expectations, we would predict a higher proportion of noticers for the semantically-related critical word compared to the phonologically-related or unrelated critical word.It is important to note that the visual world paradigm measures overt attentional shifts using eye-tracking.However, researchers have found that significant IB can persist even when a participant fixates on the critical stimulus (Koivisto et al., 2004).Considering this research within the context of the present study, future work using eye tracking could investigate whether noticing the critical word phonologically or semantically related to the primary task is associated with overt shifts of visual attention.
Future research could also use a neuroimaging study to investigate neurobiological theories of attention for the semanticcongruency effect when expectations are violated.Specifically, we are interested in communication between top-down and bottomup attentional processes attributed to the dorsal and ventral networks.This communication is associated with activation in the right temporoparietal junction (TPJ; Anticevic et al., 2010;Corbetta et al., 2000;Corbetta et al., 2008;Marois et al., 2000;Todd et al., 2005;Young et al., 2010).Our behavioural study results can be explained with reference to three theories that provide predictions for the role of the TPJ: (1) Circuit Breaker (Corbetta & Shulman, 2002); (2) Contextual Updating (Geng & Vossel, 2013); (3) Attention Schema (Graziano, 2014;2022).
The Circuit Breaker theory proposes that the TPJ is associated with attentional shifts, where activity in the region is suppressed or increased for the purpose of directing attention (Anticevic et al., 2010;Dugué et al., 2018;Emrich et al., 2011;Todd et al., 2005).In the present study, noticing the critical word could have been the outcome of two processes: suppression of activation in the TPJ while participants focused their attention on the primary task and increased activation in the TPJ to redirect attention to the critical word in the centre of the screen.According to the Circuit Breaker theory, TPJ activation is associated with reorienting attention.The predictions for both Experiments 1 and 2 will be the same: increased activation in the TPJ for participants who reoriented attention and noticed the critical word, and suppression of activation in the TPJ for participants who did not notice the critical word and remained focussed on the primary task.
The Contextual Updating theory (Geng & Vossel, 2013) and the Attention Schema theory (Graziano, 2014;2022) propose that activation in the TPJ is associated with a pre-attentive update to facilitate the allocation of attention, rather than solely for the allocation of attention.These two theories, when compared to the Circuit Breaker theory, can make more specific predictions for Experiment 1 and Experiment 2. The Contextual Updating theory proposes that the role of the TPJ is to prompt an update about the contextual environment, so as to allocate attentional resources toward meaningful stimuli.In the context of searching for insects, the critical word "BEETLE" is meaningful, and participants are more likely to notice it than the critical word "WRENCH," which is irrelevant to the task.The Attention Schema theory proposes that the role of the TPJ is to construct a schema for attention.This schema is not a direct representation of the external world but rather an internal model that continuously updates and facilitates the allocation of attentional resources (Wilterson & Graziano, 2021).In the present study, the participants' schema for attention is for whatever is relevant to the primary-task category (e.g., insects), as set by the instructions and the prior experiment trials.It is suggested that activation in the TPJ occurs when a violation to this schema prompts an adjustment to the current model.
For Experiment 1, the Contextual Updating and the Attention Schema theory provide different predictions.The Contextual Updating Theory would predict activation in the TPJ for the appearance of the critical word semantically related to the category of primary-task pictures, but not for a semantically-unrelated word.However, the Attention Schema theory would predict suppression in the TPJ when attention was directed toward semantically-related stimuli (i.e., the pictures of insects and the critical word BEETLE) because the model predicted attention for the semantic category of insects.For Experiment 2, where there was a violation to expectations for the category of primary-task pictures (i.e., participants expected pictures of insects but received pictures of tools or musical instruments), both the Contextual Updating theory and the Attention Schema theory would predict activation in the TPJ when an update about the contextual environment is required due to an unexpected change in the primary task.However, the Attention Schema theory provides the most comprehensive explanation for the results of Experiment 2. This theory predicts activation in the TPJ when there is a violation to the predictions in the schema for attention (e.g., when pictures of tools were displayed instead of insects), and there is an update to the schema for the purpose of directing attention (i.e., attention to the new semantic category).
Adapting the current experiment paradigm for use with concurrent fMRI or electroencephalogram (EEG) would create an exciting series of future studies investigating these three theories.It is acknowledged that the paradigm would require redesigning for use with these measures.For example, many more trials would be required, with multiple presentations of a critical stimulus.The dependent variable in an IB paradigm is the detection of a salient stimulus during an attention demanding task.This presents a unique challenge for the application of brain imaging techniques requiring multiple critical trials, as the fundamental design of an IB experiment involves the single presentation of a critical stimulus within a given trial.Some researchers in the field have begun this work in experiment designs such as in no-report IB paradigms (Harris et al., 2020;Hutchinson, 2019;Pitts et al., 2012;Schelonka et al., 2017;Shafto & Pitts, 2015).
While contributing novel insights into the phenomenon of inattentional blindness and attentional processes, our study is not without its limitations.One notable limitation is the absence of biological measures or brain imaging to substantiate our discussion of the role for the TPJ in attention and expectation.The study design includes only behavioural response data, and without neuroimaging data we are restricted from making strong conclusions about the role of the TPJ in this IB paradigm.Notwithstanding these limitations, the empirical results and theoretical discussion outlined in the present study do nonetheless provide a solid basis for further investigation of the role of the TPJ in attention, meaning, and expectation.

Conclusion
The present study provides insights into attention, meaning, and expectation.First, the results indicate that attention orients and reorients toward what is meaningful to us.Experiment 1 replicated the semantic-congruency effect, finding that a person's current goals guide the selection of what is attended to-when looking for pictures of insects, participants reported the critical word BEETLE that was presented unexpectedly but failed to report the critical word WRENCH when it was presented unexpectedly.Experiment 2 was the first to explore the semantic-congruency effect when participants' task expectations were violated-when the primary-task picture category was changed from insects to the unexpected picture category tools, participants reported the critical word WRENCH but failed to report the critical word VIOLIN.The findings indicate that when attention is focused, a violation of primary-task expectations alone is not sufficient to interrupt current internal processes and reorient attention.Instead, attention reorients only when the new and unexpected pictures that violated the primary-task expectations are semantically related to the critical word.This violation of task expectations prompts an update for the purpose of allocating attentional resources.This update is proposed to have included the bottom-up information about the unexpected change in picture category and the modification of attentional priorities.This reset the parameters for what was semantically relevant, and adjusted predictions for the deployment of attention.In conclusion, when attentional resources are consumed by a task, attention reorients to information that is meaningful, but what is meaningful can be quickly updated in light of unexpected information.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

S.Fig. 1 .
Fig. 1.Computer Displays with the Critical Word on the Critical, Divided-Attention, and Full-Attention Trials in Experiment 1 and Experiment 2.Note.In Experiment 1, participants were instructed to "Look for insects, ignore fruits" (Conditions 1a, 1b) or "Look for tools, ignore fruits" (Conditions 1c, 1d) and there was no violation to the primary-task picture category since participants were presented with insects and fruits (Conditions 1a, 1b) or tools and fruits (Conditions 1c, 1d).In Experiment 2, participants were also instructed to "Look for insects, ignore fruits" but there was a violation to the primary-task picture category since the category of pictures was changed either to tools and fruits (Conditions 2a, 2b) or musical instruments and fruits (Conditions 2c, 2d).

Fig. 2 .
Fig. 2. Sequence of Displays for Each Trial, Followed by Non-Critical Trials 1-4, Critical Trial 5, Divided-Attention Trial 6, and Full-Attention Trial 7.Note.The sequence of displays for each trial included a fixation cross, a primary-task display, and a blank screen.To move from the blank screen to the beginning of the next trial, participants pressed the space bar (after having given a verbal response naming the primary-task pictures, either insects or tools).A critical word appeared at fixation simultaneously with the pictures in Critical Trial 5, Divided-Attention Trial 6, and Full-Attention Trial 7.

Table 1
Experiment 1: Pattern of Task Pictures and Critical Words in Non-Critical Trials 1-4, Critical Trial 5, Divided-Attention Trial 6, and Full-Attention Trial 7.

Table 2
Experiment 1: Two-way Chi-Square Analyses Comparing Proportion of Noticers in Critical Trial 5, Divided-Attention Trial 6, and Full-Attention Trial 7.

Table 4
Experiment 2: Two-way Chi-Square Analyses Comparing Proportion of Noticers in Critical Trial 5, Divided-Attention Trial 6, and Full-Attention Trial 7.