Related Experiment Video
Updated: Aug 9, 2026

Comparing the Frequency Effect Between the Lexical Decision and Naming Tasks in Chinese
Published on: April 1, 2016
Boston naming test short forms: a comparison of previous forms with new item response theory based forms
R E Graves1, S C Bezeau, J Fogarty
1Department of Psychology, University of Victoria, Victoria, British Columbia, Canada. rgraves@uvic.ca
This study compared new and existing short forms of the Boston Naming Test (BNT) to see which ones best identify naming difficulties in patients with Alzheimer’s disease and vascular dementia. Two new versions of the BNT were developed using a statistical method called item response theory (IRT). The 30-item IRT-based form matched the full-length BNT in diagnosing abnormal naming in 44% of patients with Alzheimer’s and vascular dementia. The 15-item IRT-based form identified 48% of these patients as abnormal, with 90% agreement with the full test. An adaptive version of the test, which uses only 15 items for some patients, performed as well as the full 30-item test for 75% of patients with normal naming ability. The study shows that IRT can help create shorter, reliable versions of the BNT without losing diagnostic accuracy.
Area of Science:
- Neuropsychological assessment methods
- Cognitive disorders in aging populations
- Item response theory in psychometrics
Background:
Current neuropsychological assessments often require lengthy administrations, limiting their use in clinical settings. Prior studies have shown that the Boston Naming Test (BNT) is a widely used tool for evaluating naming abilities in patients with cognitive impairments. However, the full-length BNT may be time-consuming for clinicians. No prior work had resolved how to effectively shorten the test without losing diagnostic accuracy. This gap motivated researchers to explore alternative methods for creating shorter forms of the BNT. They aimed to ensure that the shortened versions retained high reliability and diagnostic power. The study sought to address whether item response theory (IRT) could improve short-form development. The research also aimed to compare IRT-based forms with previously developed short forms. This approach could refine diagnostic efficiency in cognitive assessments.
Purpose Of The Study:
The study aimed to develop and evaluate new short forms of the Boston Naming Test (BNT) using item response theory (IRT). Researchers focused on whether these new forms could match the diagnostic accuracy of the full-length BNT while reducing administration time. The specific problem addressed was the need for efficient yet reliable cognitive screening tools for elderly patients. The motivation came from the limitations of existing short forms, which may lack sufficient reliability or diagnostic sensitivity. The researchers proposed that IRT could better capture item difficulty and patient ability. This method could allow for adaptive testing, where fewer items are needed for some patients. The goal was to improve diagnostic efficiency without compromising accuracy. This approach could support broader use of the BNT in clinical settings.
Main Methods:
Researchers collected data from 206 elderly outpatients to develop two new short forms of the Boston Naming Test (BNT). They applied item response theory (IRT) to select items that best represent the full test's diagnostic properties. The IRT-based short forms included 30 and 15 items, respectively. They compared these new forms with 12 previously developed short forms. A sub-sample of 69 patients with mild Alzheimer's disease (AD) or AD combined with vascular dementia (VD) was used for evaluation. The researchers assessed diagnostic accuracy by comparing abnormal classifications across forms. They calculated internal consistency using Cronbach’s alpha for each form. An adaptive version of the test was also tested, which used only 15 items for 75% of patients with normal naming ability.
Main Results:
The 30-item IRT-based short form of the BNT achieved a reliability of alpha = .90 and correctly identified 44% of AD/VD patients as abnormal. This form showed 93% agreement with the full BNT in classifying these patients. The 15-item IRT-based form had a reliability of alpha = .84 and identified 48% of AD/VD patients as abnormal, with 90% agreement with the full BNT. The adaptive version of the 30/15 item test matched the performance of the full 30-item test for 75% of patients with normal naming ability. The full BNT had a reliability of alpha = .90 and identified 44% of AD/VD patients as abnormal. The IRT-based forms demonstrated comparable diagnostic accuracy to the full-length test. Both new forms outperformed previously developed short forms in terms of agreement with the full test. These results suggest that IRT can effectively guide the development of shorter, reliable cognitive assessments.
Conclusions:
The study demonstrated that item response theory (IRT) can produce short forms of the Boston Naming Test (BNT) with high diagnostic accuracy. The 30-item IRT-based form matched the full BNT in identifying abnormal naming in AD/VD patients. The 15-item form slightly improved detection rates while maintaining strong agreement with the full test. An adaptive version of the test reduced item count for 75% of patients with normal naming ability. These findings suggest that IRT-based forms may improve diagnostic efficiency without sacrificing reliability. The authors propose that these new forms could be valuable in clinical settings where time is limited. They emphasize that the IRT approach allows for better item selection based on difficulty and patient ability. The study supports the use of IRT in developing neuropsychological assessments.
Frequently Asked Questions
The 30-item IRT-based short form matched the full BNT in identifying abnormal naming in AD/VD patients with 93% agreement.
The 15-item form identified 48% of AD/VD patients as abnormal, with 90% agreement with the full BNT.
IRT allows for better item selection based on difficulty and patient ability, improving diagnostic accuracy in shorter assessments.
The adaptive version matches the performance of the full test but requires only 15 items for 75% of patients with normal naming ability.
The 30-item IRT-based form had a reliability of alpha = .90, matching the full BNT.
The authors proposed that IRT is a valuable method for developing shorter, reliable neuropsychological assessment tools.
Related Concept Videos
Multiple Comparison Tests
It would be easy to compare two samples using a significance alpha level of 0.05. In other words, there is only one sample pair to be compared. However, it would be difficult to identify a significantly different sample if the number...
Comparing Experimental Results: Student's t-Test
Binet's Contribution to Measures of Intelligence

