• Raw Scores
• T-scores and Percentiles
• Confidence Intervals

Raw Scores

Raw scores are derived by summing the numeric values of the item responses that were endorsed by a rater on a particular form for a given scale. For most of the items, response options are defined so that higher scores indicate more marked symptoms or impairment. For example, a Hyperactivity raw score of 28 indicates a considerably higher level of reported hyperactive behaviors than does a raw score of 4. However, there are items where the response options represent the severity of symptoms or impairment in the other direction (i.e., lower scores indicate more marked symptoms or impairment). These items are reverse-scored when calculating a raw score, which means that the value of the response is inverted. For example, responses range from 0 to 3; if a response of 3 is provided for a given item, the reverse-scored value is 0, and the resulting item score of 0 contributes to the calculation of the scale’s raw score.

Beyond generic interpretations, however, one person’s raw score on a scale is not readily comparable to those received by other people, and a person’s various scale scores are not comparable to each other because they contain a different number of items. To facilitate interpretation of the results, raw scores are converted to standardized scores, namely T-scores and percentiles.

T-scores and Percentiles

A T-score is a standardized score that can be easily compared across different raters or administrations. T-scores for Conners 4 scales convert the raw scores to reflect what is typical or atypical for a particular age (and gender, if Gender Specific reference samples are selected). All T-scores have a mean of 50 and a standard deviation of 10, and as such, allow for meaningful comparisons between the youth being evaluated and those in the reference samples selected.

Table 4.1 provides guidelines for understanding Conners 4 standard scores. This table provides terms that can be used to describe the standard scores. For example, a T-score in the “Very Elevated” range (i.e., greater than or equal to 2 standard deviations above the mean of the reference sample selected) for a certain scale suggests that the rater (i.e., the parent, teacher, or the youth) has reported greater or more significant problems about the youth’s functioning in this specific area than what is typically reported for youth of that age (and gender, if Gender Specific reference samples are selected). T-scores below 60 generally indicate typical levels of symptoms or functioning, or fewer or no more problems than typically reported; that is, the rater reports the youth is functioning at a level that is not atypical compared to others of the same age (and gender, if using Gender Specific reference samples). At the lower end, the Conners 4 scores have been capped at a T-score of 10, and at the upper end, the scores were capped at a T-score of 100 (i.e., T-scores below 10 will appear as 10, and T-scores greater than 100 will appear as 100 in the reports).

There are two important things to highlight. First, a T-score in the “Slightly Elevated” range requires careful consideration and clinical judgment, as this range is the borderline between the high end of the average range and the lower end of the elevated range. In such cases, it is especially beneficial to consider other information, such as confidence intervals, to determine which category the case is more likely to be. Second, scores that border two categories warrant special consideration. For example, T-scores of 69 and 70 fall in different categories (“Elevated” and “Very Elevated,” respectively), even though there is no significant difference between these scores. It may be helpful to describe a T-score of 69 or 70, for example, as falling on the border of the “Elevated” and “Very Elevated” categories.

Scores on the Conners 4 are also described as percentiles, which are the empirical percentile ranks based on responses from the selected reference sample(s). Empirical percentiles were selected to represent the true shape of the distribution of responses and are determined using the observed cumulative frequency counts of the selected reference sample(s). Percentiles indicate the percentage of the youth in the same age group (and gender, if using Gender Specific reference samples) in the selected reference sample(s) who obtained a raw score that is the same as, or lower than, that obtained by the youth who was rated. For example, a score at the 90th percentile is equal to or greater than 90% of the scores for that scale in the reference sample selected. Percentiles can be used to compare an individual to their peers and can offer a useful way to explain this comparison to an individual who does not understand T-scores.

T-scores and percentiles are provided for the Conners 4 Content, Impairment & Functional Outcome, and DSM Symptom Scales. Note that the guidelines shown in Table 4.1 are approximations and should not be used as absolute rules. Clinical training and judgment are required for responsible interpretation of any test score; these score classification guidelines should not be applied automatically without careful interpretation by a qualified user.

Confidence Intervals

All measurements contain error (see Naglieri & Chambers, 2009, for a discussion of measurement error) and this should be acknowledged whenever scores are presented and compared. Interpreting results while ignoring measurement error may produce misleading conclusions for the user (Oosterwijk et al., 2019; Wilkinson & Task Force on Statistical Inference, 1999).

Measurement error in Conners 4 scores is represented with confidence intervals. Confidence intervals provide a range of T-scores, at a particular level of probability, within which the individual’s true score could be expected to occur (Harvill, 1991). That is, if an individual was evaluated 100 times and a 95% confidence interval was created each time, then 95 of those 100 confidence intervals would be expected to contain their true score. The width of the confidence interval indicates the precision of the estimate; more confidence can be placed in estimates with narrow confidence intervals than in those with wide confidence intervals (Morris & Lobsenz, 2000; see Standard Error of Measurement in chapter 8, Reliability, for guidelines). A less reliable scale score (one with a greater error in measurement) will have a wider confidence interval than more reliable scores. The level of confidence selected will ultimately depend on the user’s comfort level. Most commonly, 90% and 95% confidence levels are used. Accordingly, both levels are available for Conners 4 T-scores, calculated based on each scale’s standard error of measurement (SEM); note that the 90% confidence interval is the default setting for reports. See chapter 3, Scoring and Reports, for more details on selecting the level of the confidence interval and appendix B for tables with values of the confidence intervals for all T-scores by scale for all Conners 4 forms. Further details about measurement error in the Conners 4 can be found in Standard Error of Measurement in chapter 8, Reliability.