Running Head: WARNING: REPEATING TABOO AND NEGATIVE WORDS IN RSVP CAN CAUSE BLINDNESS.

Warning: Repeating Taboo and Negative words in RSVP can Cause Blindness.
Perception and Cognition

Abstract
The effect of emotional salience on perceptual processing was explored using the RSVP experiment paradigm at stimulus speeds of 100ms/frame. 24 students completed 80 trials of three word streams, each containing two target words. Target words were identical for the ID condition and different for the control condition. The words were divided into taboo, negative emotion, positive emotion and neutral words. Results showed that all ID categories showed more Repetition Blindness (RB) than the control condition and that taboo and negative emotion words generated more RB than neutral and positive words. Interestingly, neutral words demonstrated a higher RB than positive emotion words, possibly suggesting the salience of negativity over emotion.

Method
Participants
A total of 24 students comprised of 15 females and 9 males, from an undergraduate second year psychology class took part in the experiment. The mean age was 28.8, and the ages ranged from 19-50. All participants first learned to read and write in English, and 21 also spoke English as their first language.
Materials
Participants used the lab computers to complete the trials. Each trial consisted of three words, which included the target words C1 and C2 and a filler word in between them. The target words were identical in the treatment group and different for the control group. There were four types of words used for the four different conditions: taboo; negative emotion; positive emotion; and neutral. Participants completed a total of 80 trials each and the words used for the different conditions can be found in Appendix A.
Procedure
Participants completed the trials as part of their psychology tutorial and had the opportunity to opt out of the experiment if they did not wish to read the explicit taboo words. Participants completed an online demographic survey before beginning the experiment. They were given five practice trials and were provided with a numbered and lined A4 page to write down the words they saw after each trial. Each trial began with a centred plus symbol, followed by a masking frame of centred characters, the first target word C1, the filler word, the second target word C2 and a prompt to provide time for participants to write down the words, before the participants initiated the next trial. Each frame lasted for 100ms.

Results
Results indicated that the control condition word category means were significantly higher than their corresponding category means in the identity condition. The mean difference can be viewed to represent the difference in the level of Repetition Blindness (RB) for each category. The mean differences for each category were ranked in the following descending order: taboo; negative; neutral; and positive. The mean difference for taboo words was .6609 - .0217 = .6392, and was significant at t(23) = 25.63, p = .000. The mean difference for negative words was .6391 - .0435 = .5956, and was significant at t(23) = 19.55, p = .000. The next highest mean difference was for the neutral words category with a mean difference of .6217- .0565 = .5652. It was significant at t(23) = 19.80, p= .000. The lowest mean difference .6478 - .2652 = .3826 occurred for the positive words category at t(23) = 9.67, p= .000.
Discussion
The original hypothesis predicted that identity RB would decrease in the following order of experimental conditions: taboo; negative; positive; and neutral. As expected, all conditions in the identity condition had significantly different means to the control condition means. Experimental results only partially supported the predicted order of emotionally salient words producing a stronger RB effect than the neutral condition, with taboo words demonstrating more RB than negative words, and both of them exhibiting higher RB than neutral and positive words. However, the results did not support the expectation that positive words would have a higher RB than neutral words. Instead, neutral words were found to have a higher RB than the positive valence condition, the only emotional salient condition that had a lower RB than the neutral condition. The possible reasons for the unexpected result will be explored later in the discussion.
Taboo and negative words generated the highest identity RB effect respectively, therefore partially agreeing with the assertion that a stronger activation, relative to neutral words, can result from the inherent emotional salience of the stimuli (Silvert, Naveteur, Honore, Sequeira & Boucart (2004). The Silver et al (2004) study only analysed negative emotion words, whereas our experiment explored the effects of taboo and positive emotion words as well. The positive emotion words were found not to display a higher RB relative to neutral words.
Therefore, the current experiment added to research by demonstrating that taboo words have a stronger activation than negative, neutral and positive words respectively, due to their larger resultant RB. However, the differences between the conditions were not tested for significance. The results also indicated that in the type-token process (Kanwisher, 1987), the token individuation phase for taboo words is stronger than the other conditions. Moreover, the results also confirmed previous findings that negative emotion words would be more activated, and subsequently display higher RB than positive emotion words (Nasrallah, Carmel & Lavie 2009). According to the type-token account (Kanwisher, 1987), this is the reason why it is more difficult for participants in the taboo and negative conditions to recall the second instance of the repeated target word than the participants in the neutral and positive conditions.
These results differ from the MacKay, Hadley and Schwartz (2005) study that investigated orthographic RB for taboo and neutral words. Their study surprisingly found that taboo words generated lower orthographic RB relative to neutral words. However, their divergent results must be examined in light of the many differences in research design, methodology and definitions present in their study, compared to the common practices for empirical investigation into RB.
The present study’s results also added to the research by clarifying the importance of negative emotion words relative to positive emotion words in predicting RB magnitude. It showed that negative valence emotion words were more difficult to recall than positive valence emotion words as hypothesised. This supports the findings by Nasrallah, Carmel and Lavie (2009), where negative emotion words were more sensitive to perception and categorisation than positive emotion words, when extreme valency and arousal were controlled. It further shows that negative emotion words have a stronger activation (Kanwisher, 1987) than positive emotion words.
This study’s hypothesis that positive emotion words would be more strongly activated than neutral words was proposed because positive emotion words were considered to be more associated with the effect of being in an emotional category as a result of their positive valence than neutral words were (Silvert, Naveteur, Honore, Sequeira & Boucart, 2004). Research in perceptual processing under RSVP conditions to date has predominantly focused upon comparing the processing of emotion words and in particular negative emotion words relative to neutral words, with no direct empirical comparison of positive emotion words to neutral words under RSVP conditions.
For this experiment, the positive emotion words were not expected to produce less RB than the neutral condition. Even though there were no tests conducted to examine whether the differences were significant between them, the positive emotion conditions repeated word recall (26.52%) was much higher than the neutral (5.65%), negative (4.35%) and taboo words (2.17%). It appears the positive emotion words category was the most different to the other categories. This difference cannot be explained conclusively due to the preliminary nature of the experiment. Also, the lack of previous research makes it difficult to discern with certainty, whether the hypothesis was partially incorrect, or whether the experiment’s methodology can explain the unexpected results. Further research needs to directly compare these two conditions, in the form of more sample positive words and more participants, to determine whether this assumed connection of emotionality between negative and positive emotion words is empirically justified, or whether stronger token individuation (Kanwisher, 1987) is predominantly applicable to negative emotion words and taboo words, rather than positive emotion words (Silvert, Naveteur, Honore, Sequeira & Boucart, 2004).
This experiment’s methodology may have influenced the surprising results in a number of ways. Firstly, the experiment had small word sets for each emotional category, making it more susceptible to being influenced by the participant’s interpretation of the particular words selected rather than their objective emotional category (Nasrallah, Carmel and Lavie, 2009). Secondly, and possibly more saliently, the pilot study sample may have been too small as well. This makes the mean scores generated for each word’s valence and arousal less valid a representation of the entire sample. This claim is supported by the anomaly that arose in the experiment when analysing the mean valence ratings of the pilot study, where taboo words were expected to register more negative ratings than negative words, due to the obscene aspect of their meaning. Instead, using a Likert scale, the mean valence for the taboo words was 2.20, a score more positive or less negative than the negative words mean valence of 1.40. The unexpected ratings may have been significantly influenced by outlier scores for certain words, as there were no control for extreme valence scores, as in previous studies (Nasrallah, Carmel and Lavie (2009). Subsequently, the pilot study’s positive emotion words mean valence rating of 4.15 and neutral rating of 3.11 may not be as truly representative of the individual words scores in their category due to the way they were influenced by extreme scores. Previous studies have averaged out each word valence to reduce this limitation (MacKay, Hadley & Schwartz, 2005).
Similarly, when considering the mean arousal ratings generated for each condition, the relative difference in magnitude between the negative and taboo conditions, 3.95 and 3.10 respectively, to the neutral condition of 2.92, appear to be significantly different to the difference between the mean arousal for the positive condition of 3.03. However, this difference has not been confirmed statistically. The lack of difference in arousal between positive emotion words and neutral words could suggest that arousal may be interacting with the valence to affect the results. A bigger difference in arousal between the positive and neutral category, to parallel the difference seen in the negative and taboo word category with respect to neutral words, may have resulted in more emotional type and initial token individuation (Kanwisher, 1987) for the positive condition and greater RB overall (Anderson, 2005). This further creates a doubt surrounding the assumption that arousal and valence can be considered separate dimensions, particularly since taboo words may have high arousal purely by definition. The relationship may be more related regarding the valence of the word.
The nature of the experiment also allows for the possible explanation that positive emotion words produce less RB than neutral words. The results also allow space for speculation, since positive emotion words have not been directly explored relative to neutral words previously. Using the evolutionary and adaptability theory arguments that are usually applied to explain the perceptual sensitivity towards negative emotion words and applying them to positive emotion words, it may be speculated that positive words are activated the least out of the word categories because they are deemed to be the least harmful and threatening. When conceptualised as being threatening or non-threatening, positive emotion words appear further away from negative words than neutral words. Therefore, neutral words may be treated more suspiciously and require more time for processing than positive words because their valence is closer to the negative valence than the positive emotion words. Conceptually speaking, the perceptual system may categorise stimulus as being negative or non-negative, rather than emotional or neutral.
One of the main limitations of the experiment concerned the limited statistical treatment of the mean results for the conditions in the treatment group and the pilot study. This made it difficult to reach conclusive findings as the level of significance between means was not investigated. There were also methodological limitations such as having a low sample of words for each condition (Nasrallah, Carmel & Lavie 2009) and not adequately controlling for arousal and the effects of extreme valence ratings. Individual arousal and valence ratings for the whole sample would be beneficial in the future to gain a better reflection of the arousal and valence of individual words.
As previously mentioned, future research would benefit from a more explicit comparison between positive emotion words and neutral words under RSVP conditions. This experiment also highlighted the issue of determining valence for taboo and negative words and whether cultural changes are changing the expected valences of different categories of emotional words. Similarly, the overlap between arousal and valence needs to be further explored in both negative emotion and positive emotion words.
In conclusion, the preliminary findings of the experiment suggest negative emotional salience has an effect on perceptual processing. Taboo and negative emotion words generate a stronger identity RB effect than neutral and positive emotion words. The experiment also revealed a probable significant difference in processing between negative emotion words and positive emotion words, possibly suggesting the salience of threat and arousal in the processing of emotional words.

References
Anderson, A.K. (2005) Affective influences on the attentional dynamics supporting awareness. Journal of Experimental Psychology: General, 134, 258-261.
Kanwisher, N. G. (1987). Repetition blindness: Type recognition without token individuation. Cognition, 27, 117–143.
Kanwisher, N. G., & Potter, M. (1990). Repetition blindness: Levels of processing. Journal of Experimental Psychology: Human Perception and Performance, 16, 30–47.
MacKay, D. G., Hadley, C. B., & Schwartz, J. H. (2005). Relations between emotion, illusory word perception, and orthographic repetition blindness: Tests of binding theory. The Quarterly Journal of Experimental Psychology, 58A(8), 1514-1533.
Nasrallah, M., Carmel, D., & Lavie, N. (2009). Murder, she wrote: Enhanced sensitivity to negative word valence. Emotion, 9(5), 609-618.
Silvert, L., Naveteur, J., Honoré, J., Sequeira, H., & Boucart, M. (2004). Emotion stimuli in rapid serial visual presentation. Visual Cognition, 11(4), 433-460.