The present study investigated object-based feature encoding in visual short-term memory for 2 features within the same dimension that occur on different parts of an object. Using the change-detection paradigm, this experiment studied objects with 2 colors and objects with 2 orientations. Participants found it easier to monitor 1 rather than both features of such objects, even when decision noise was properly controlled for. However, no object-based benefit was observed for encoding the 2 features of each object that were of the same dimension. When similar stimuli were used but the 2 features of each object were from different dimensions (color and orientation), an object-based benefit was observed. These results thus impose a major constraint on object-based feature encoding theories by showing that only features from different dimensions can benefit from object-based encoding.How are features of objects perceived and retained? In an early study by Allport (1971), colored numerals inside colored shapes were presented briefly for recall of one or more features. It was found that report of a form feature (e.g., the shapes) was not affected by whether a color feature (colors of the shapes or numerals) was also reported, but was negatively affected by the report of another form feature (e.g., the numerals). A second study by Wing and Allport (1972) confirmed these findings and found that the report of both spatial frequency and orientation did not affect either report, but the report of two orientation features (grating orientation and an orientation of a transverse break in the lines of the grating) interfered with each report significantly. These authors argued that perceptual analysis occurs in systems of analyzers, each dealing with a specific feature dimension such as color or orientation. As such, two features from different dimensions can be encoded in parallel without mutual interference, but two features from the same dimension will have to share the same analyzer and cannot be encoded without interference (see also Treisman, 1969). Duncan (1984) questioned the above conclusion and argued that feature grouping by objects plays a more important role in perception. He asked participants to report two of four features from four independent dimensions (size, tilt, texture, and the location of a gap), either located on the same object or on different objects. He found that two features located on the same object could be reported more readily than two features located on different objects. In a later study in which two letters were presented, Duncan (1993) asked participants to report (a) the size and shape of one letter, (b) the same attribute (size or shape) of two letters, (c) the shape of one letter and size of the other, or (d) both attributes of both letters. He found that report of (a) was much better than that of (b), (c), or (d) and that the differences among (b), (c), and (d) were not significant. In a second experiment, Duncan (1993) presented two objects, each containing orientation, length, and fr...