Study: How Typos Affect Perception of Intelligence

This study investigates whether grammatical errors in social media posts affect perceptions of authors' intelligence. Using a randomized controlled experiment with 265 participants from Amazon Mechanical Turk and UC Berkeley XLab, we examined the impact of two types of spelling errors: typographical errors (accidental keystroke mistakes) and phonological errors (spelling based on word sounds).

Our findings reveal statistically significant negative effects for both error types on perceived intelligence. Phonological errors resulted in a 1.14-point decrease on a 7-point Likert scale (p < 0.01), while typographical errors caused a 0.56-point decrease (p < 0.01). These results suggest that readers make unconscious judgments about authors' cognitive abilities based on spelling accuracy, with sound-based spelling errors being particularly damaging to perceived intelligence.

Command of language is one of the most significant cognitive abilities we possess and is often the most pervasive signal we encounter in a social media setting. When we notice overt and unintentional grammatical errors in social media posts, do we make unconscious assumptions about the authors' general intelligence? Do we attribute difficulty with written language with other indicators such as lower-performing verbal acuity or overall intelligence? Further, are some categories of grammatical errors more injurious than others -- or do we take in stride all these trespasses?

General intelligence, sometimes referred to as cognitive ability, includes our capacity to "reason, plan, solve problems, think abstractly, comprehend complex ideas, learn quickly, and learn from experience" (Plomin, 1999). Assessment of our cognitive abilities often informs critical judgments in others that affect our educational, occupational, and relationship opportunities.

Though social media channels are often used to identify potentially qualified job candidates, their use in screening candidates for suitability and background investigation is also on the rise. In a CareerBuilder survey, 57% of employers reported rejecting candidates based on negative findings in applicant social media posts. Of those rejected, 27% specified "poor communication skills" as the primary factor for the rejection.

Our hypothesis is that both typographical and phonological spelling errors, compared to no spelling errors at all, will lead to a decreased level of perceived intelligence, irrespective of the nature, content, and platform of the social media post.

There is evidence that we ought to take this question seriously. In Borkenau & Liebler (1993), college students rated their perceived intelligence of strangers after watching them read aloud a pre-written weather report. The study found a significant correlation between perceived and measured IQ scores of the strangers, suggesting that some information about individual intelligence is provided through verbal communication.

Kreiner et al. (2002) showed that, while experiments with only a small percentage of typographical errors didn't result in significant perceived intelligence ratings, the presence of a larger number of typographical errors or phonological errors did significantly influence the perception of cognitive writing abilities.

The distinction between error types is crucial. Typographical errors (typos) are spelling mistakes that occur due to slip of the hand when typing, including artifacts such as swapping letters, typing a letter close to the intended letter on the keyboard, or leaving out a letter. Usually the misspelled word is still easy to recognize (e.g., "somethimg", "hwo", "thrd").

Phonological errors, on the other hand, arise out of ignorance, or spellings based on sounds of the word. These usually demonstrate that the author truly does not know the proper spelling of the word (e.g., "sumthing", "Kansus", "razing").

Experimental Design

We adopted a pre-test/post-test control group study design. Participants were randomized at the start of the study with equal probability of assignment to Control, Typographical, or Phonological groups. All participants first viewed one identical post containing no errors to establish a pretreatment baseline. Each group then viewed 5 additional posts specific to their treatment condition.

Our null hypothesis states that individuals exposed to either treatment will not view its author as less intelligent than those exposed to control. Our alternative hypothesis proposes that treatment exposure leads to decreased perceived intelligence ratings.

Measurement

Participants were invited to join a short, anonymous survey with the stated intent of assessing their opinion on the appropriateness in length of social media posts. Deception was employed to avoid anticipated response bias. For each social media post, participants answered:

• Attention Question: Recall question about the post content
• Decoy Questions: 5-point Likert on post length, 7-point Likert on attitude toward content
• Outcome Questions: 7-point Likert scales for perceived effectiveness, intelligence, and writing skills
• Error Detection: Question asking how many spelling/grammar mistakes were noticed

Pilot Study

Our pilot was conducted March 10-11, 2021, with open solicitations posted on Facebook and UC Berkeley's Slack. We received participation from 31 volunteers.

Despite the small pilot dataset, we observed highly significant results: phonological errors had more than double the effect of typographical errors on perceived intelligence.

Live Study

The live study was conducted March 31 - April 9, 2021, with participants recruited through Amazon Mechanical Turk (March 31 - April 1) and UC Berkeley XLab (April 6-9). We collected data from 265 participants.

Participant Demographics

Our study drew participants primarily from the United States, with a diverse distribution across gender and age groups.

Response Quality Analysis

We analyzed response quality by examining reading speed (words per minute) and Likert scale response variance to identify potential low-effort respondents.

Placebo Test

To assess data quality, we performed placebo tests measuring treatment effect on the pretreatment control question. The Amazon MTurk data showed suspicious significance, suggesting potential randomization issues. The XLab data passed the placebo test, so we used only XLab participants (N=209) for our primary analysis.

Baseline Model

Our baseline model examines only the effect of treatment on perceived intelligence using robust standard errors. Both treatment effects are negative and highly statistically significant.

The phonological treatment shows a 1.142-point decrease in perceived intelligence, while typographical errors show a 0.558-point decrease. Phonological errors have approximately twice the negative impact of typographical errors.

Model with Demographic Controls

Adding demographic covariates (gender, age, education, native language, social media usage) slightly increases the magnitude of treatment effects while maintaining statistical significance.

Treatment Effect Visualization

The divergence plots below show the distribution of Likert scale responses by treatment group. Bars extending left represent lower ratings (1-3), while bars extending right represent higher ratings (5-7).

Compliance Analysis

We analyzed potential noncompliance using an instrumental variables approach. Compliers were defined as participants who: (1) correctly identified error presence/absence in posts, and (2) had reading speeds between 50-500 words per minute, suggesting careful reading.

Complier rates: 78% Phonological, 76% Typographical, 63% Control.

Interpretation

Our results strongly support the hypothesis that grammatical errors in social media posts negatively affect perceptions of author intelligence. The effect is robust across multiple model specifications and analytical approaches.

The finding that phonological errors have approximately twice the impact of typographical errors is particularly noteworthy. This may reflect that phonological errors signal a fundamental lack of word knowledge, while typographical errors can be attributed to carelessness or typing speed.

Implications

These findings have practical implications for social media users, particularly in professional contexts:

• Careful proofreading before posting may help maintain positive impressions
• Phonological errors should be particularly avoided as they carry greater reputational cost
• Employers using social media screening should be aware of potential biases in intelligence attribution

Limitations

• Sample composition: Primarily UC Berkeley students, limiting generalizability
• Artificial setting: Participants knew they were being studied, potentially affecting behavior
• Post content: We used neutral topics; effects may differ for emotionally charged content
• Platform specificity: Results may not generalize to all social media platforms
• MTurk data quality: Failed placebo tests required excluding MTurk participants

This randomized controlled experiment provides strong evidence that grammatical errors in social media posts causally affect perceptions of author intelligence. Both typographical and phonological errors result in statistically significant decreases in perceived intelligence, with phonological errors having approximately twice the negative impact.

Our findings extend previous research by confirming these effects in a contemporary social media context using rigorous experimental methodology. The results suggest that language errors serve as powerful signals that readers use to make inferences about cognitive ability, even in informal online communication.

Future research should examine whether these effects persist across different demographic groups, cultural contexts, and social media platforms. Additionally, investigating potential interventions to reduce bias in intelligence perception based on writing errors would be valuable.

1. Borkenau, P., & Liebler, A. (1993). Convergence of stranger ratings of personality and intelligence with self-ratings, partner ratings, and measured intelligence. Journal of Personality and Social Psychology, 65(3), 546-553.
2. Driver, S. (2020). Keep It Clean: Social Media Screenings Gain in Popularity. Business News Daily. Link
3. Kreiner, D. S., Schnakenberg, S. D., Green, A. G., Costello, M. J., & McClin, A. F. (2002). Effects of spelling errors on the perception of writers. The Journal of General Psychology, 129(1), 5-17.
4. Newman, B. M., & Newman, P. R. (2020). Theories of Adolescent Development. Academic Press.
5. Plomin, R. (1999). Genetics and general cognitive ability. Nature, 402(6761), C25-C29.
6. Silverman, F. H. (1990). Are professors overrating students who stutter? Journal of Fluency Disorders, 15(2), 81-85.