The QWERTY Effect: How Typing Shapes Word Meanings and Baby Names- Casasanto, Jasmin, Brookshire & Gijssels

Introduction Some form-meaning relationships in language may be arbitrary (de Saussure 1996). However, some word meanings are constrained by the way they are produced. Ohala (1984)- systematic relationship between form and meaning of words denoting small vs large. Chico/gordo (Spanish), petit/grand (French), mikros/makros (Greek). In these cases, the small word requires the speaker to shorten the vocal tract, and a vowel in the large word requires the speaker to lengthen it.. Spoken words are increasingly being produced with the hands as we type, mediated by the QWERTY keyboard. Allows for new form-meaning relationships in language to arise. On average, words types with more letters from the right side of the QWERTY keyboard are more positive in meaning that words types with more letters from the left. This is the QWERTY effect, shown in English, Spanish and Dutch in a large corpus of phonotactically legal English pseudowords and in a collection of keyboard-based neologisms. Was predicted on the basis of a more general relationship between left-right space and emotional valence. Casanto (2009)- people tend to associate positive with their dominant side, and negative with their non-dom side. So for right-handers, right = good, left = bad. This implicit association is enshrined in idioms like my right hand man and two left feet found across many languages- majority of language users are right-handed. From this, can be predicted that words typed with right-side are more positive in meaning. Experiments 1-2: QWERTY effect in a large English corpus and extended to Portuguese. Experiment 3- extended to German and ruled out one of the possible origins of QWERTY effect previously proposed. Experiment 4- relationship between keyboard positions and valence extends to individual letters. Left-handers show a similar QWERTY effect to right-handers, further constraining theories about the origin of the QWERTY effect. Experiment 4- showed a strong relationship between QWERTY key position and baby-names, suggesting a real-world impact. Experiment 1: QWERTY in larger corpusMethod and Results Analysed words from NewANEW corpus (Warriner et al 2013), 13915 words normed for Valence, Arousal and Dominance. Subjects rated words on these dimensions using a 9-point scale. Calculated the right side advantage for each word by taking the difference of the number of the letters on the right side [LSA] of the keyboard and subtracting the number from the left side [RSA]. Only the Valence scores were of interest, so Arousal and Dominance scores were not analysed. 3 analysis: 1- relationship between Valence and RSA in simple linear regression. 2- relationship between Valence and RSA controlling word length, letter frequency and their interaction. 3- relationship between Valence and RSA using a nonparametric randomisation test- provides an intuitive way to assess the observed relationship. In first simple analysis- valence was positively related to the RSA, but the relationship was not signficiant. In word length analysis- RSA was a significant predictor of Valence. In randomisation test, there was a less than 1/100 chance of obtaining relationship between RSA and Valence. In summary- English words with more right-side letters were on average more positive in valence than left-side words.Experiment 2: QWERTY Effect in PortugueseMethod and Results Analysed valence-normed words from the EP-ANEW corpus (Soares et al 2001). 1034 words, rated by 958 native Portuguese speakers for Valence, Arousal and Dominance, using 9 point scales. Diacritics were stripped from the letters, because, on the Portuguese keyboard, accented letters do not have their own keys. 19 words had hyphens, treated as right-side letters. In simple regression, RSA was highly significant in Valence. In word length control- showed predicted positive relationship, though the effect was marginally significant. In summary: In the controlled analysis, the marginal relationship is likely due to relatively small number of words in the corpus. To determine whether this effect differed in magnitude from English, Spanish and Dutch, a regression analysis combining data from all four languages was made, with translated equivalents as a repeated random factor. The magnitude of QWERTY effect did not differ between languages- indicated by the absence of any RSA by Language interaction.Experiment 3: QWERTY Effect in German Tested whether key position predicted valence for German words, in order to extend the QWERTY effect to another language. Also tested the possible origin of the QWERTY effect that initially predicted that the effect on the basis of manual motor asymmetries in right-handers- should prefer right-hand letters because they are easier to type with the dominant hand. There is an asymmetry built into QWERTY keyboards- there are more left-handed letters than right-handed letter, creating more response competition for letters typed with the left hand than the right. However, the QWERTY keyboard has an equal number of right/left hand letters- testing for a QWERTY effect allows us to determine whether the left-right symmetry that gives rise to the QWERTY effect is located in the keyboard or in the body of its right-handed users.Method and Results Analysed valence-normed words from BAWL-R corpus (Vo et al 2009). Consists of 2902 words rated for valence, arousal and imageability. Valance ratings made of a 7-point scale from -3 to +3. Only valence ratings were analysed. In simple linear regression, RSA was a significant predictor of Valence. In word length control, RSA was a highly significant factor. In summary: QWERTY effect was highly significant in both raw and controlled analysis, extending the effect of RSA onto a 5th language. This argues against the possibility of locating the origin of the effect in the bodies rather than the keyboard.Experiment 4: QWERTY Effect in Letters Tested for effect of QWERTY key position on English speakers valence rating for individual letters, compared this effect between right and left handers.Method and Results Native English speakers recruited via Amazon Mechnical Turk, 209 participated for online payment. Of these, 7 were excluded for not following instructions, and 4 excluded for having non-QWERTY keyboard, so 198 remaining. Each letter from the alphabet was presented in lower case on an individual page, with order randomised. Participants indicated how positive the letter seemed on a 9 point SAM scale. Each letter was assigned a value corresponding to its left-to-right column position on the keyboard. In simple regression- keyboard column was a significant predictor of valence. Same positive relationship when controlling for a letters frequency, its ordinal position in the alphabet and their interaction. Tested whether the relationship between keyboard column and valence was categorical or continuous. This relationship could be categorical- letters typed with the left hand could be treated as more negatively and those typed with right hand could be positive. Alternatively, this relationship could be continuous- increase gradually column-by-column from left to right. According to the non-parametric rank-order test of the relationship, the relation appears to be continuous. Finally, added Handedness and its interaction to the model. Measured by the Edinburgh Handedness Inventory (EHI: Oldfield 1971). Right-handedness was operationalised as having a EHI greater than 40 (168 right handers) and left-handedness had EHI less that -40 (10 left handers, 20 ambidextrous excluded from analysis). The effect of horizontal position did not differ between right/left handers- showed similar QWERTY effects in tests of the effect of Keyboard Column on Valence, controlling for the letters Ordinal Position in the alphabet, their frequency and their interaction between original position and frequency. Overall, the effect of keyboard column on valence remained significant when handedness, the interaction of handedness with keyboard column, the letters ordinal position in the alphabet, their frequency and interaction between ordinal position and frequency were all controlled. In summary: The relationship between keyboard position and valence was found in individual letters- letters father to the right were rated as more positive in valence than letters of the left. Was found in spite of the fact that a located in the left-most column was the most frequently positively rated letter by far- presumably because A signifies top rating in school grades and other evaluation schemes. Meaningfulness of this letter and its status works against the right-is-good relationship, which was found in both right/left handers.Experiment 5: QWERTY and Baby Names Although the effect is subtle, it is pervasive- may shade the meaning of words with every letter we type, read and imagine. One domain of language use in which people have a greater degree of autonomous choice, and therefore where QWERTY effect is likely to be found, is in naming new places, products of people. Tested whether first names that Americans give their children have changed over time, as QWERTY has become more of a norm in language usage, and whether names coined after QWERTY are spelled more right-sided.Method and Results Naming statistics from the US Social Security Administration website. Reports the frequency of each name that was given to five or more children in the US that year. Longitudinal analysis of names popularity: First analysed the mean RSA of all names from 1960-2012 that had been given to at least 100 children every year- 788 names. Results showed that the mean RSA has increased since the popularisation of the QWERTY keyboard, indicated by correlation between the year and average RSA. Remained highly significant when word length was controlled by dividing RSA by the number of letters in a name. Comparison of pre-versus post-QWERTY era names: May also influence how new names are coined. RSA of names coined before and after QWERTY popularisation was compared. Is difficult to pinpoint the moment in history when QWERTY became common and part of daily lives. Apple Mac and Windows home became available in 1984/5 but were not widely used. American Online made the internet widely available in 1991. Chose 1990 as the beginning of the QWERTY era based on a survey of technological landmarked like those listen above. Names invented after 1990 (38,746) use more letters form the right side of the keyboard than pre-1990 (43,429). This difference remained significant when length was controlled by dividing each names RSA by the number of letters in the name. In summary, very strong relationships are found between names Americans have decided to give their children and the QWERTY key positions of the letters in those names. In longitudinal analysis, we found that the mean RSA of the names that were already in use as of 1960 increased dramatically starting the dawn of the QWERTY era- higher-RSA names were increasing in popularity. In second analysis, found that names coined after 1990 have higher RSAs than names used during the previous 3 decades.General Discussion First proposed that left-right asymmetry that gives rise to the QWERTY effect could either be inherent in the bodies of the right-handed users or in the keyboard. Finding a QWERTY-like effect in German argues strongly against this second possibility. Whereas the QWERTY keyboard has more left-handed letters than right letters, the German QWERTZ does not- since it cannot be explained by asymmetries in the keyboard, it would be inaccurate to build this into the theory of other languages. The relationship between keyboard position and the word meanings in all of the languages tested, therefore, is best explained in terms of manual motor asymmetries inherent to the users, the great majority of whom are right handers. QWERTY effect in single letters rules out two other possible explanations for the effect: Typing letter combinations that use a) distinct fingers or b) alternating hands is easier than typing combinations that reuse the same fingers or the same hand (Beilock & Hock 2007). Could the QWERTY effect be driven by some unexpected relationship between the right-left positions of keys and finger repetitions or hand alternations? Previously, J&C ruled out effects of FRs and Has statistically, showing that the effect of RSA on valence remained significant when these variables were controlled- we can rule out FRs and Has definitively- the relationship cannot be explained in terms of typing easy vs hard key combinations. Finally, theorising about the effects origins is strongly constrained by the finding of a significant QWERTY effect in left handers, which did not differ quantitatively or qualitatively from the effect found in the right-handers. This replicated previous findings that showed a trend toward the standard right-biased QWERTY effect in left-handers. On the simplest prediction, RHs should show a RS advantage, while LH show LS advantage. This follows from studies showing that LHs manifest a good is left bias in numerous ways, rooted in LH motor fluency (Casasanto 2009). Both LH and RH prefer words with more right-side letters suggests that the QWERTY effect arises both from typing experience and from the experience of using words in speech. Variation in the way words are used is constrained by communities of language users, which are composed mostly of RH-ers. English speakers must all agree that the correct answer is the right answer, even if they are LHers. Perhaps good is associated with right on the keyboard because good is associated with right, more generally, in a social word dominated by right-handers.