Quantitative text analysis in gifted and talented research

ABSTRACT

The authors introduce readers to quantitative text analysis and its potential for analyzing text-based nonreactive measures as a means of broadening the evidential basis within research on giftedness and talent. After defining quantitative text analysis and describing how it can augment survey studies for analyses at individual and cultural levels, the special potential of quantitative text analysis for addressing cultural questions within gifted and talented research is described. Inductive and deductive techniques of quantitative text analysis are then introduced and illustrated with brief examples drawn from gifted and talented research. Finally, a brief illustrative study is presented via an extended series of linked worked examples focusing on the question of the place of giftedness and talent within mentoring discourse in US mass media of past and present in order to illustrate in more detail how quantitative text analysis can be used to improve the culture-level evidence base within gifted and talented education.

KEYWORDS:

• Gifted and talented
• gifted education
• talent development
• mentoring
• quantitative text analysis
• nonreactive measures

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. Other nonreactive measures such as physical traces of human behavior can also overcome response bias and reference-group effects (Fritsche et al., 2006). However, they lie beyond the remit of our article on quantitative text analysis.

2. The asterisk is a wildcard denoting the inclusion of all word forms (e.g. mentor, mentors, mentored, mentoring, mentorship).

3. Luhn (1960) proposed 60-character passages for display—for efficient visual inspection—in portrait layout on 8.5-by-11-inch paper. Our example passages in Table 2 are about twice as long, containing 107.5 characters on average (SD = 11.61), but intended for display on landscape layout, which ensures that each passage still fits in one line and can be quickly reviewed.

4. One rater rated privilege and privileged as not qualifying as at least a loose partial synonym of any of the four target concept terms. However, the rater subsequently agreed that the two terms do qualify as partial synonyms of the target concept terms. For this reason, we retained privileg* in our final dictionary.

5. While some recommendations for assessing the interrater reliability of lookup dictionaries advise using a stricter adjustment for chance agreement provided by Cohen’s κ (for an overview, see Rau & Shih, 2021), the variance among the raters was too low to allow for calculating Cohen’s κ, and questions have recently arisen about the appropriateness of using Cohen’s κ for interrater assessments of nominal data (Rau & Shih, 2021).