A critical assessment of potential measurement biases in the technology acceptance model: Three experiments

The Technology Acceptance Model (TAM) is widely used by researchers and practitioners to predict and explain user acceptance of information technologies. TAM models system usage intentions and behavior as a function of perceived usefulness and perceived ease of use. The original scales for measuring the TAM constructs have been confirmed to be reliable and valid in several replications and applications spanning a range of technologies and user populations. However, a measurement bias may be present because the TAM instrument physically groups together the multiple items measuring each individual construct. Many scholars of instrument design would object to such item grouping, instead advocating that items from different constructs be intermixed in order to reduce "carryover" effects among the responses to multiple items targeting a specific construct, which might artificially inflate the observed reliability and validity. Three experiments involving two systems and a total of 708 subjects are reported which address whether such carryover biases are present in the TAM measures. All three studies found that item grouping vs. item intermixing had no significant effect (positive or negative) either on the high levels of reliability and validity of the TAM scales, or on the path coefficients linking them together. Ironically, open-ended verbal evaluations indicated that subjects were more confused and annoyed when items were intermixed, suggesting a tendency toward "output interference" effects, which themselves could have a biasing effect. Our findings suggest that those who employ the TAM measures should continue using the original (grouped) format in order to best predict and explain user acceptance of information technology.