Perception & Psychophysics 1985, 38 (3), 261-268 Perceptual bias and response bias in temporal bisection THOMAS G, RASLEAR Walter Reed Army Institute ofResearch, Washington, DC In a temporal bisection task with rats, perceptual bias and response bias were simultaneously varied through manipulations of stimulus spacing and the relative probability of reinforcement for correct responses. Both manipulations produced systematic changes in the bisection point. However, only manipulations of relative reinforcement probability produced significant variations in B ", a nonparametric index of response bias. This finding shows that the bisection point may be shifted by either a perceptual bias or a response bias. However, in the absence of an index of response bias, such as B", shifts caused by perceptual effects are indistinguishable from those caused by response preferences. Temporal bisection is employed as a behavioral tool to animal analogue of the human bisection task was first detect the effects of drugs on time perception (Maricq & described by Boakes (1969), and subsequent research has Church, 1983; Maricq, Roberts, & Church, 1981; Meck, generally followed his procedures with some minor var­ 1983). Changes in the bisection point indicate whether iations.' drugs and other treatments affect the perception of time An animal bisection experiment has two distinct phases: directly, by changing the speed of a hypothesized internal discrimination training and generalization testing. In the clock, or indirectly, by modifying memory processes. training phase, subjects learn a simple two-choice dis­ However, an important aspect of drug effects on behavior crimination between the two stimuli (A and B) that define has not yet been evaluated for the bisection task: it is not the sensory interval to be bisected. Responses on known whether response bias may influence the bisection manipulandum A are reinforced on trials in which point, and thus cause response preferences to be mistaken stimulus A is presented, and responses on manipu­ for true perceptual effects. This is particularly important landum B are reinforced on trials in which stimulus B is with respect to some of the drugs and treatments that have presented. Inappropriate responses are never reinforced. been used in the bisection task to date. For instance, am­ Once an accurate discriminative performance is achieved, phetamine can produce response stereotypies (Iverson & the proportion of correct responses that are reinforced is Iverson, 1981), inescapable footshock often produces a gradually reduced to the level that will be present during depression of movement (Anisman, de Catanzaro, & generalization testing. Remington, 1978), and response perseveration may follow During generalization testing, several "test" stimuli, septal or hippocampal lesions (Braggio & Ellen, 1976; intermediate in value to the training stimuli, are in­ Ellen, Makohon, & Richardson, 1978). If any of these troduced. Responses in the presence of the "test" stimuli effects were present in the bisection task, a response are never reinforced, whereas correct responses in the preference or bias could change the bisection point and presence of the training stimuli continue to be reinforced be mistaken for a true perceptual effect. at a level that maintains the same overall probability of The bisection task as implemented for animal subjects reinforcement as existed during the final stage of the is different in several respects from the procedure used training phase. Responses made in the presence of each with human subjects. Human subjects require little, if any, of the stimuli (test and training) are recorded and used formal training in the task, whereas extensive training is to construct a generalization gradient. The generalization required to ensure that animal subjects are attending to gradient is usually the proportion of response A (or B) the stimulus dimension of interest and will perform the that occurred as a function of stimulus magnitude. The response(s) from which a bisection point is derived. An proportion of A responses is near 1.0 in the presence of stimulus A and near 0.0 in the presence of stimulus B. The intermediate-value "test" stimuli produce proportions This material has been reviewed by the Walter Reed Army Institute that lie between the two extremes. The bisection point of Research, and there is no objection to its presentation and or publi­ (BP) is commonly taken as the stimulus magnitude that cation. The opinions or assertions contained herein are the private views produces a 0.5 proportion of A responses. of the author and are not to be construed as official or as reflecting the As noted above, the potential role of response bias in view of the Department of the Army or the Department of Defense. the animal bisection task has not yet been evaluated. Requests for reprints should be sent to Thomas G. Raslear, Department of Medical Neurosciences. Division of Neuropsychiatry. Walter Reed However, it is known that the bisection task, as im­ Army Institute of Research, Washington, DC 20307-5100. plemented for humans and animals, is subject to context 261 262 RASLEAR effects or perceptual bias. For example, changes in 100 stimulus context, such as stimulus spacing, stimulus fre­ A ...... quency, or stimulus order in time and space, have been en ..... LU shown to affect psychophysical tasks in humans (Helson, en PERCEPTUAL BIAJ.··· z 1964). Fagot and Stewart (1970) demonstrated that the 0 CL. BP shifts with changes in stimulus spatial order in a human en t········ LU brightness bisection task, and Raslear has demonstrated a:: 50 NORMAL that changes in stimulus spacing produce shifts in the BP zt:l .' 0 for rats in auditory bisection (Raslear, 1975) and in -J temporal bisection (Raslear, 1983) tasks. #. The demonstration that manipulations of stimulus context affect the outcome of a psychophysical task can be interpreted in at least two ways, as Helson (1971, p. 6) 0 has noted: .. Adaptation-level theorists maintain that with 0 5 10 changes in stimulation accompanying focal stimuli the TIME (sec) quality, magnitude, and other dimensions of the stimuli also change more or less. Critics of AL theory assert that the changes under such conditions reported by the subjects 100 do not reflect changes in perceived attributes but are B merely verbal, semantic, or judgmental artifacts." en RESPONSE BIAS....·· Response preferences are the "verbal, semantic, and judg­ LU enz J......... mental artifacts" of concern in an animal bisection task, 0 so the issue is whether changes in stimulus context produce enCL. .' LU 50 changes in the BP because the perceived magnitudes of a:: .' NORMAL the stimuli have been affected (i.e., perceptual bias) or zt:l because of an induced response preference (i.e., response 0 -J bias). #. Most human studies of perceptual bias have not ad­ dressed the issue of whether such biases may be explained in terms of response bias as defined by signal detection 0 theory (however, see Hellstrom, 1985). It is possible that 0 5 10 manipulations of stimulus frequency merely act to move TIME (sec) the criterion in animal as well as in human studies. Changes in stimulus spacing may merely produce changes 100 in the relative reinforcement rates for the two responses in the bisection task. Since responses to the test stimuli C are not reinforced in the typical animal bisection pro­ en cedure, and some ofthe test stimuli are likely to be poorly LUen z discriminated from one of the training stimuli, a smaller 0 CL. proportion of choices on one response lever would be rein­ en LU 50 forced. This would result in a decrease in responses on a:: .' c::l .' that lever and a shift in the BP. The shift would not be Z .' 0 .' a true perceptual effect, but an instance of response bias. -J Figure I provides a schematic illustration of how per­ #. .' ceptual bias, response bias, and a change in sensitivity .' would influence the psychometric function in a bisection 0 task. The normal psychometric functions are represented 0 5 as straight lines for ease of presentation. As shown in 10 panel A of Figure I, a pure perceptual effect (i.e., TIME (sec) response bias and sensitivity are unchanged, but the BP has shifted) would cause a bowing ofthe normal psycho­ metric function without any simultaneous changes in the Figure 1. Illustrations of the effects of perceptual bias, response two anchor or training stimuli. A pure response bias effect bias, and sensitivity on the psychometric function for bisection. In (only sensitivity is unchanged) would cause an upward each case, the normal psychometric function is shown as a straight line for ease of presentation. Percent "long" responses are plotted (or downward) shift in all stimuli, as shown in panel B as a function of linear time, and changes in the bisection point (50% of Figure 1. Finally, a pure change in sensitivity (BP and "long" responses) are shown. BIAS AND BISECTION 263 response bias unchanged) would be seen as a change in terminated the trial and, with a predetermined probability, produced slope of the function, as seen in panel C. a 45-mg Bio-Serv food pellet, whereas a single incorrect response Although several studies have demonstrated shifts in merely terminated the trial. Ifno response was made within 10 sec, the BP with changes in stimulus context (Raslear, 1975, the trial terminated and a null response to that stimulus was recorded. The intertrial interval was 10 sec, during which responses had no 1983), no study has shown that manipulations of response effect. Each stimulus occurred with a probability of .5 on each trial. bias can also affect the BP. The experiment described here A session consisted of 300 trials, of which the first 20 were used addressed this question by simultaneously varying stimu­ as a warm-up for the animals, with a probability 1.0 ofreinforcement lus spacing and the relative probability of reinforcement for correct responses.