The visibility of composite stimuli (double lines, multiple lines and sinusoidal gratings) is analysed in terms of the visibility of single lines. 2. The detection mechanisms for the composite stimuli were examined by measuring the probabilities for seeing these stimuli, the probabilites of "recognition" near their detection thresholds and by using the method of subthreshold summation. 3. The simplest composite stimulus consists of two lines spaced apart by 15 min; it was found that both the probability of seeing this stimulus and the probability of reporting either one or two lines may largely be explained by assuming independent detection of the two lines. For a 12 min spacing some antagonistic interaction was observed between the two lines, confirming the existence of an antagonistic surround in the line detector. 4. The probability of seeing a multiple line target (rectangular grating) of spacing 15 min is slightly greater than that predicted assuming independent detection of its component lines; this probability can be calculated from the probabilities of seeing a central line (near the fixation point) and pairs of lines spaced 15, 30, 45, 60, 75 and 90 min on either side of the fixation point. The observed threshold for the grating was about 0.05 log units less than the prediction. 5. The probability that a subject reports only one line from a rectangular grating was found to be consistent with independent detection of its component lines when the subject used a low (detection) threshold criterion but not for higher criteria. 6. The spatial frequencies involved in the detection of the rectangular grating were determined by subthreshold summation; the high spatial frequency selectivity was shown to be consistent with independent detection of the component lines. 7. The detection mechanism for a 4 c/deg sinusoidal grating was found to be similar to that for the 15 min rectangular grating. 8. Evidence is presented that the detection of finer and coarser gratings may be analysed in similar terms. 9. It is shown that lines are not detected by independent activation of sharply tuned grating detectors but that sub-units may be involved which are slightly more narrowly tuned than the line detector. 10. These sub-units contain both inhibitory and disinhibitory regions; the disinhibitory responses may be the basis of the deviations from probability summation noted in 4. 11. The properties of the proposed sub-units are in reasonable accord with the results of adaptation experiments.
