Marine Biology (1994) 120:161-169 9 Springer-Verlag 1994 C. G. Lowe 9R. N. Bray 9 D. R. Nelson Feeding and associated electrical behavior of the Pacific electric ray Torpedo californica in the field Received: 26 January 1994 / Accepted: 27 January 1994 Abstract The predatory behavior of 74 Pacific electric rays (Torpedo californica), studied between August and December during 1988 through 1991 in situ off the Palos Verdes Peninsula, southern California, consisted of two feeding modes: an ambush from the substratum during the day and a more vigorous attack from the water column at night. Predatory motor patterns and electric organ discharges (EODs) were recorded on the video and audio channels of a housed camcorder. Predatory motor patterns included four phases: (1) jump (simultaneous with EOD initiation), (2) pectoral-fin cupping, (3) orientation to prey, and (4) ingestion. The initial electrical activity of the rays was a train of 46 to 414 5-ms monophasic EODs that lasted 0.45 to 7.06 s; the maximum number of EODs produced during an attack was >1200. Maximum output, measured on only one ray, was 45 V. Fifty-five rays were presented one of four types of prey stimuli: live fish (LF), freshlyspeared fish (FSF), frozen fish (FF), or a simulated bioelectric field (SBF). The percent frequency of attacks for the LF, FSF, and FF treatments ranged from 70 to >90%, but was <30% for the SBF. The interval between prey presentation and attack was =30 s for the LF, FSF, and FF and over five times longer for the SBF; intervals averaged <4 s for the three rays tested at night. Attacks by rays on energized electrodes provide the first evidence that electric rays use electroreceptors to detect their prey. However, the lack of clear differences among the four prey treatments in five characteristics of the initial pulse train suggests that a suite of sensory stimuli cooperate in triggering an attack and regulating the electrical output during the attack. Communicated by M. Strathmann, Friday Harbor C. G. Lowe1 9R. N. Bray2 ([]) 9D. R. Nelson Department of Biological Sciences, California State University, Long Beach, California 90840-3702, USA Present addresses: 1 University of Hawaii at Manoa, Department of Zoology, Honolulu, Hawaii 96822, USA 2 Biology Program, California State University, San Marcos, California 92096, USA Introduction Electric rays (family Torpedinidae) are one of only two groups of strongly electric fishes found in the marine environment (Bennett 1971). Studies to date indicate that rays of the genus Torpedo are piscivorous and produce electric discharges while attacking prey (e.g. Bray and Hixon 1978). Attacks from the substratum in aquarium observations typically start with a "jump" over the prey and a simultaneous train of EODs (Belbenoit 1970, 1986, Belbenoit and Bauer 1972). The predatory behavior varies considerably, presumably due to differences in the chain of events during an attack, stimuli emitted by prey, and age of the ray (Belbenoit 1986). An added complication of laboratory observations, however, is that stress incurred during capture and maintenance and the artificial nature of the laboratory environment may affect the behavior and electrical activity of the rays (e.g. Bennett et al. 1961; Pfeiffer 1961). The only field observations on the predatory behavior of electric rays indicate that they actively seek prey at night (Bray and Hixon 1978). Diver surveys over a rocky reef in southern California indicated that Pacific electric rays (Torpedo californica) were far more common swimming above the reef at night than during the day. Rays aggressively attacked and consumed speared fish presented to them, and Bray and Hixon concluded that the rays were actively hunting for reef fish exposed but quiescent at night. A tracking study of Pacific electric rays using ultrasonic telemetry provides additional evidence that Pacific electric rays swim about mainly at night (Bray, Lowe, and Nelson in preparation). The purpose of this study was to quantify electrical activity during predatory attacks of previously undisturbed Pacific electric rays in the field. Field measurements of the electrical activity of electric rays have not been reported and we were interested in whether measurements made from laboratory studies were conservative. Additionally, since electric rays may feed mainly at night or in turbid water (Pfeiffer 1961; Bray and Hixon 1978), we were inter- 162 e s t e d in h o w t h e y d e t e c t and e v a l u a t e t h e i r prey. W e varied the n a t u r e o f p r e y s t i m u l i b y p r e s e n t i n g r a y s w i t h either l i v e , f r e s h l y - s p e a r e d , or f r o z e n f i s h or a s i m u l a t e d bio e l e c t r i c field, and r e c o r d e d the s u b s e q u e n t m o t o r p a t t e r n s and e l e c t r i c a l a c t i v i t y o f the rays. Materials and methods Our study site was Abalone Cove at Palos Verdes Peninsula in southern California (33~ 118~ The cove consists of a cobble beach, a subtidal rocky reef with attached giant kelp (Macrocystis pyrifera), and an offshore sand flat. We made most of our observations over the sand flat, beginning at the outer margin of the reef at a depth of 8 to 12 m and extending to a depth of = 23 m. Rays (Torpedo californica) were in 10 to 15 m of water and usually buried in the sand during the day and swimming in the water column at night. Tests were conducted between August and December during 1988 through 1991. After testing, each ray was measured (total length) to the nearest 10 cm using a calibrated pole, sexed, and tagged with a color-coded spaghetti dart tag; the tagging ensured that each ray was tested only once. EOD waveform measurements The waveforms of the electric organ discharge (EOD) produced by the rays during predatory events were recorded with a Tektronics model 214 storage oscilloscope in a watertight housing. We used two types of recording electrodes for measuring EODs underwater: a fixed dipole, with positive and negative poles spaced 20 cm apart, to measure the voltage of the electric fields created by the ray's EODs; and an adjustable caliper dipole, clamped directly on the dorsal and ventral surfaces of the electric organ to measure the maximum voltage output of rays. Both electrodes were stainless steel rods, 2 mm in diameter. The voltage drop through the electrodes was negligible when in direct contact with a known voltage source in a seawater tank. We conducted tests only on previously untested and undisturbed rays found buried in the sand during the day. Male and female rays between 60 and 110 cm were presented with live and freshly-speared reef fish ranging from 15 to 20 cm (standard length, SL). Speared fish, impaled on the positive electrode pole, were presented within 10 cm of the anterior margin of the ray's disk and were moved back and forth in front of the ray until a response was elicited. Following each response, waveform information (amplitude, polarity, EOD duration, and EOD shape) was transcribed underwater from the oscilloscope screen to a slate. Table 1 Torpedo californica. Stimulus characteristics of the four different prey treatments (P present; P§ present at increased strength). Elevated electrical strength of FSF based on research by Kalmijn (1972) Prey treatment EOD pulse-train measurements We recorded the electrical and behavioral patterns produced by rays with an underwater video camera (a housed Sony CCD-V3 8 mm camcorder). The system was fitted with a monopole electrode connected through a voltage splitter to the audio channel, thus providing a simultaneous record of rays' movements (video) and EODs (audio) during a predatory attack. For each test, the monopole recording electrode was placed =10 to 20 cm above the dorsal surface of an electric organ just before the prey stimuli were presented. We used four different treatments representing various prey stimuli to study how rays detect their prey (Table 1). Previously undisturbed male and female rays were randomly presented either live (LF), freshly-speared (FSF), or frozen then thawed (FF) fish, all measuring 15 to 20 cm standard length and commonly found around rocky reefs in southern California. These species (Rhacochilus vacca, Embiotoca jacksoni, Phanerodon furcatus: Embiotocidae; Halichoeres semicinctus, Oxyjulis californica: Labridae; Chromis punctipinnis: Pomacentridae; and Medialuna caIiforniensis: Kyphosidae) were either caught by hook and line and lightly anesthetized with MS-222 (LF treatment) or speared with a pole spear (FSF treatment); the fish used during a particular trial depended on its immediate availability. For the LF treatments, each fish was tethered with 20-poundtest monofilament nylon line to a 1 m plastic rod, allowing us to move the prey to within 10 cm of the anterior margin of the ray's disk. We also examined the ray's response to a simulated bioelectric field (SBF). The SBF generator, modified from a design used by Kalmijn (1978), produced an 8 gA DC current across a 5 cm dipole. The apparatus had two sets of salt-bridged dipole electrodes (stainless steel, 1.5 mm diam), 50 cm apart, and a three-way switch. During a trial, only one set of electrodes, randomly chosen, was energized. This enables us to control for the visual and mechanical stimuli also produced by a moving electrode. The audio portion of each test was printed out on a Gould Mark 220 high-speed oscillographic chart recorder and then digitized with a digitizing tablet for measurement of time intervals between individual EODs. From these intervals, we determined total train length and total number of EODs within the train. We limited some analyses to the initial part of the total sequence of EODs because of the likelihood of experimenter-induced artifacts during subsequent EODs. The end of the initial pulse train was defined as the first 80 ms gap between consecutive EODs. We computed instantaneous pulse rates as the reciprocal of the time interval between adjacent EODs; from these data, we then computed the average and maximum instantaneous pulse rate within the initial train. Since diving logistics resulted in small sample sizes and unbalanced design, which could result in inaccurate probability (p)-values, we compared the p-values of parametric analyses with those computed from comparable randomization tests using 10 000 reshufflings (Sokal and Rohlf 198I, Crowley 1992); the two p-values always agreed within 0.01. Results Response Visual Mechanical Electrical Chemical voluntary a twitch Live fish (LF) P Freshly-speared fish (FSF) P Frozen fish (FF) P Simulated bioelectric field (SBF) P P P P P P+ P+ P a Swimming movements of body and fins A total o f 74 Torpedo californica w e r e t e s t e d b e t w e e n A u g u s t and D e c e m b e r d u r i n g 1988 t h r o u g h 1991. F e m a l e rays were significantly longer than males (females: s = 88_+12 c m (+1 S D ) ; m a l e s : 2 = 7 9 + 6 c m ; S t u d e n t ' s t-test t = - 3 . 3 4 , d f = 7 2 , p = 0 . 0 0 3 ) , but w e r e less c o m m o n (19 f e m a l e s , 55 m a l e s ; G - t e s t G = 18.28, df= 1, p < 0 . 0 0 1 ) . EOD waveform characteristics E O D w a v e f o r m s w e r e r e c o r d e d f r o m 11 o f the 13 rays that w e r e p r e s e n t e d f r e s h l y - s p e a r e d p r e y (the o t h e r t w o r a y s 163 did not attack). Individual EODs consisted of a DC pulse with a rapid rise to peak and slow decay to baseline; the dorsal surface of the electric organ was always positive and the ventral surface negative. Pulse durations ranged from 4 to 5 ms (Y = 5 +0.5 ms). The amplitude of the waveforms varied greatly, ranging from 0.02 to 0.50 V cm -1 (~ = 0.17 _+0.18 V cm -1 , n = 11). There was little correlation between EOD amplitude and total lengths of rays ( r = - 0 . 1 0 , p = 0.768). The large variation in amplitude measurements is at least in part due to our difficulty in placing the dipole electrodes at a consistent distance and orientation from the rays. Only one ray was tested for m a x i m u m electric organ output using the caliper electrodes. The m a x i m u m voltage, measured by clamping the caliper electrodes on the dorsal and ventral surfaces of the ray's electric organ, was 45 V, with a waveform similar to those observed in the other tests. Predatory motor patterns Daytime predatory motor patterns of rays showed stereotyped chains of behaviors similar to those described by Belbenoit ( 1981). Of the 10 ray s presented live tethered prey, 8 were initially buried in sand (e.g. Fig. 1 A, C) and two were swimming (e.g. Fig. 1 B, G). Predatory behaviors started with a jump over the prey (Fig. 1 D) (simultaneous with initiation of the EODs), followed by pectoral-fin cupping around the prey (Fig. 1 E, H), orientation of the ray so that it could grasp the fish head first (Fig. 1 F, I - J ) , and final ingestion of the prey whole. All 10 rays presented with live fish initiated their attacks with a jump over the prey (Fig. 2); however, 4 of the rays missed or apparently lost sensory contact with the prey after the initial jump (Fig. 2 e, f, g, j). Pectoral-fin cupping and orientation behavior immediately followed a jump in 9 of the 10 cases, with durations depending on the time required for manipulation of the prey for ingestion. Two of the ten rays apparently had difficulties in handling prey because of interference with the tether; they stopped their attacks and swam away without ingesting the fish (Fig. 2 a, h). In one instance, we inadvertently presented two prey to one ray (Fig. 2j). The closer fish, a halfmoon (Kyphosidae: Medialuna californiensis), was still subdued because of the anesthetic while the farther fish, a black perch (Embiotocidae: Embiotocajacksoni), was alert and swimming from the ray's left to right side. The ray jumped toward the black perch, turning to its right as the fish passed. At the beginning of the jump, the ray produced a burst of EODs that immobilized the black perch (Fig. 2j: Ja); however, the ray over-jumped the fish and ceased the attack. The ray jumped 20 s later toward the halfmoon (J2), just missing it. The ray jumped again 03) and landed on the immobilized black perch, ingesting it 01) after 30 s of pectoral-fin cupping. The ray then attacked the subdued halfmoon with a slow jump (J4). The fish was accidentally pulled away because the tether line became tangled. The ray then produced the most intense EOD burst observed, followed the fish into the water column, and ingested it (I2). The ray al- Fig. 1 Torpedo californica. Positions and motor patterns of Pacific electric rays before and during daytime predatory attacks. A Ray buried in sand with its head to right. B ray swimming within 1 m of bottom; white-tipped rod in diver's left hand is electrode connected with a wire to audio channel of the camcorder, black-tipped rod in right hand was used to position a tethered fish (beneath ray's disk, out of sight). C - F predatory sequence of 80 cm male ray, initially buried in sand, attacking a 17 cm standard length (SL) blacksmith (Chromispunctipinnis; Pomacentridae); C prior to attack, ray's head was directly facing camera and prey swam in from the ray's left; D - E jump with a simultaneous turn to ray's left using thrust from tail fin; F orientation to prey and ingestion; timings after initiation of attack were D = 0.3 s, E = 0.5 s, F = 0.8 s. G - J predatory sequence of 70 cm male ray, initially swimming 2 to 5 cm above bottom, attacking a 21 cm SL black perch (Embiotoca jacksoni; Embiotocidae); H late jump and early envelopment stages; I - J envelopment and orientation; timings after initiation of attack were H=l.8 s, I=2.6 s, J=3.4 s (from 8 mm video frames) 164 Fig. 2 Torpedocalifornica. Pulse profiles and associated motor patterns of ten Pacific electric rays presented live fish on a tether; rays are ordered by increasing time length of predatory attack. Symbols above pulse profiles represent time course of motor events, where J(1-4) = a jump over the prey, M = ray missed prey during jump, horizontal bar = duration of pectoral-fin cupping and orientation, and I(t - 2) = point of ingestion (EODs electric organ discharges) If I 200 - J z" -~M( J2 100 I 200 100 I ,0 C \ I 200 i JJ ~100 I oLtJ I 0 4, I :3 a. 100 0 J1 200 L]l I ,L I le ..... 0 I_ _1 10 20 . . I.. 30 . ..[ 4O . _ Ii 50 s 60 I $ 0 J I I 10 20 30 z.O 50 s 60 J1 J2 200 0 I jJzt -~ 100 100 I ~J P 200 ~,~--J3 12 I M[ 0 ways directed its jump toward the most active fish. It took the ray =2 min to attack and consume both fishes. In all trials, the initial EOD occurred at the beginning of the jump and the discharge train usually continued into the pectoral-fin cupping phase (e.g. Fig. 2). Of the 35 rays that attacked prey, 22 continued to discharge at a slower and more variable rate during the orientation and ingestion phases than during the jump phase; 13 stopped the attack 40 Time 8O tL s 120 during various portions of the first three phases and either swam away immediately or stopped searching for the fish after several minutes. The frequency of EODs at the start of the trains was typically 150 to 200 EODs s-1 and d e creased over time (Fig. 2). We observed the same predatory attack phases during the three night tests except that the rays were either swimming or drifting in the water column 1 to 2 m above the 165 substratum. Rays lunged forward at prey with a motor pattern similar to a jump and then cupped their pectoral fins around the prey. At this point, rays would either pin the fish to the bottom or go into a spiral or forward-rolling maneuver, depending on the ray's distance above the substratum; these behaviors led to orientation and ingestion of the prey. Additional details of predatory behavior of electric rays at night are described in Bray and Hixon (1978). 100 \ E 8 We compared the time interval between prey presentation and attack for the 36 rays that attacked prey stimuli (LF: 10, FSF: 15, FF: 7, SBF: 4). There was a significant difference in latency time among the four prey treatments (one-way ANOVA; F=15.30; df--3,32; p<0.001). Although rays responded to the three fish treatments at approximately the same mean latency, rays took five times longer to respond to the SBF treatment (Fig. 3 b). Since previous observations by Bray and Hixon (1978) suggested that rays actively seek prey at night, we predicted that latency time would be lower at night. Only three Ingestions L 50 o 2oo c o W IO0 Frequency of attacks and ingestions Latency of attack ~ 13_ Comparison among prey stimuli We presented one of the four prey treatments to 55 rays that ranged in total length from 60 to 100 cm (Y=79.9+ 7.9 cm). The number of rays receiving the LF, FSF, FF, and SBF stimuli was 11, 19, 10, and 15, respectively. The total lengths of rays among the four stimuli groups were similar (one-way ANOVA); F - 1 . 0 3 , df=3,49; p=0.388). Thirty-six of the rays attacked the stimuli within 4 rain of presentation; the remaining 19 did not attack and eventually swam away. Water temperatures during the experiments ranged from 13 to 18 ~ (Y = 15.4+ 1.9C ~ and differed significantly among the four prey treatments (oneway ANOVA; F = 4.47, df= 3,22; p = 0.014). Water temperatures were lower during the FSF presentations (Y= 13.3 ~ than the three other treatments [SBF: 15.6, LF: 15.9, FF: 16.3 ~ Tukey's honest-significant difference (HSD) test, p < 0.05]. The frequency of attacks differed significantly among the four prey stimuli (G-test, G = 15.03, df=3, p=0.002; Fig. 3 a). The attack percentage for the SBF treatment was <30% while those for the other treatments ranged from 70 (FF) to 91% (LF); when the SBF treatment was removed from the analysis, the differences in attack frequency among the three remaining treatments were not significant (G= 1.56, df= 2, p = 0.459). Of the rays that attacked, the frequency of ingestions also differed significantly among the three fish stimuli (omitting SBF; G-- 10.4, df=2, p= 0.006; Fig. 3 a). Ingestion percentage was lowest for FF (14%) and was identical for LF and FSF (80%). All three rays presented fish prey at night responded with a predatory attack; of these, two rays ingested the fish. ~Attacks oi 6 ~q 0 G &oo 0 W d 200 Z o % ,9, 120 80 4O 'r If 300 200 O LU loo LF, FSF FF Treatments SBF Fig. 3 Torpedocalifornica. Summary of predatory attacks and associated electrical activities of rays receiving 1 of 4 prey stimuli (LF live fish; FSF = freshly-speared fish; FF frozen fish, SBF simulated bioelectric field), a Percentage of attacks out of all rays presented treatments (openbars) and percentage of ingestions out of all rays that attacked (hatched bars); b latency between stimulus presentation and jump; c - f are for initial pulse train only: c duration, d number of EODs, e average pulse rate, f maximum instantaneous pulse rate. Bars represent means + 1 SD rays were tested at night with the monopole electrode (one with live fish and two with freshly-speared fish), so we pooled daytime LF and FSF treatments to allow for a d a y - night comparison of latency times. The rays did respond more quickly at night (2 = 3.6 + 1.2 s, n = 3) than during the day (Y =28.9+28.4 s, n = 25); these differences in latency were statistically significant [Student's t-test, t= 2.52, df= 26, p (one-tailed) = 0.009]. 166 Pulse-train characteristics We obtained usable pulse-train data on 26 rays (LF: 10 rays, FSF: 6, FF: 6, SBF: 4). Total lengths of these rays ranged from 70 to 100 cm (Y=81.2+7.1 cm) and did not differ significantly among the four treatments (one-way ANOVA; F = 1.66; df= 3,22; p = 0.204). The total number of EODs produced during predatory attacks ranged from 46 to 1234 (y = 282.6 +231.3 EODs) and occurred over intervals ranging from 0.45 to 128.2 s. Some of this variation was due to experimenter influence. For example, we observed that the duration of the entire pulse train could be lengthened if the diver continued to move the tethered fish during an attack. The initial pulse train, however, was unaffected by these artifacts. As a result, we restricted our comparison among treatments to the characteristics of the initial pulse trains during an attack. The length of the initial pulse trains ranged from 0.45 to 7.06 s and averaged 3.46 + 1.86 s. Although initial pulse trains for FSF were nearly twice as long as those for LF, FF, and SBF, the differences were not significant, possibly because of the large within-group variation (Fig. 3 c, Table 2). The total number of EODs within the initial train ranged from 45 to 414, averaging 163.8+ 105.2 EODs. The FSF received more EODs per pulse train than the other treat- Table 2 Torpedo californica. Results of one-way ANOVAs comparing five characteristics of initial pulse trains among LF, FSF, FF, and SBF prey stimuli. The dfin all five tests were 3,22 (EODs electric organ discharges) Property of initial pulse train F p Length Number of EODs Average instantaneous pulse rate Decline in frequency of EODs Maximum instantaneous pulse rate a 1.34 1.01 2.94 0.42 3.12 0.286 0.406 0.056 0.744 0.047 a Data were log-transformed to meet assumptions of homogeneity of variances Table 3 Torpedocalifornica. Partialcorrelations between five characteristics of initial pulse trains during predatory attacks by 26 rays presented 1 of 4 prey treatments Train length No. of EODs 0.95 ** Average instantaneous pulse rate -0.59"* Decline in frequency of EODs 0.77 ** Maximum instantaneous pulse rate 0.43 * No. of EODs Average Decline in instantaneous frequency pulse rate of EODs -0.45" ments, but the differences were not significant (Fig. 3 d, Table 2). The average instantaneous pulse rate within the initial pulse trains ranged from 60.8 to 122.4 EODs s -1, averaging 90.2+ 16.7 EODs s-1 among the four treatments. The LF elicited the highest mean pulse rate and the FSF received the lowest (Fig. 3 e); the differences in pulse rates among treatments were marginally significant (Table 2). All four treatments showed a general decline in frequency of EODs throughout the course of the initial trains (e.g. Fig. 2 for LF only). The least-squares slopes of the regression between pulse rate and elapsed time were all negative and did not differ significantly among the four treatments (Table 2). The maximum instantaneous pulse rate ranged from 129 to 308 EODs s-1 and averaged 178.4+_42.7 EODs s -1. The LF and FF mean maximum pulse rates were similar and substantially higher than the FSF and SBF treatments (Fig. 3 f). The differences in maximum pulse rates among treatments were marginally significant (Table 2); however, an a posteriori test was unable to identify similar groups (Tukey's HSD, p > 0.05). Within the LF treatments, there was a significant positive regression between average instantaneous pulse rate and water temperature (average instantaneous pulse rate --4.9 + 6.61 9 temperature; r 2 = 0.62; p = 0.007). Additionally, maximum instantaneous pulse rates were higher at higher water temperatures (maximum instantaneous pulse rate = 20.2 + 11.1 9 temperature; r 2 = 0.30; p = 0.10), and the relationship probably would have been statistically significant with a larger sample size. Among all four treatments taken together, however, there was no relationship between any of the four pulse characteristics and either water temperature or total length of the rays (ANOVAs, p-values for overall slopes of temperature and size covariates > 0.05). The five characteristics of the initial pulse train (train length, number of EODs, average and maximum instantaneous pulse rates, and decline of the pulse rate over time) were intercorrelated. Partial correlation analysis indicated that the train length was positively correlated with number of EODs, decline in frequency of EODs, and maximum instantaneous pulse rate, but was negatively correlated with average instantaneous pulse rate (Table 3). The number of EODs was positively correlated with decline in frequency of EODs, and was negatively correlated with average instantaneous pulse rate. Finally, average instantaneous pulse rate was negatively correlated with the decline in frequency of EODs (Table 3). When all five characteristics were compared simultaneously among the four treatments, the overall differences among treatments were nearly significant (MANOVA; Wilk's lambda=0.33; F = 1 . 6 9 ; df= 15,50; p = 0.084). 0.78 ** -0.73 ** Discussion 0.34 0.00 * Significant at p <_0.05; ** significant at p < 0.01 0.20 This study is the first to record the electrical activity of electric rays in the field. Our aim was to study the electri- 167 cal activity and behavior of undisturbed rays as they attacked fish prey. Since the behavior of rays in captivity may not reflect natural behavior, it is important to compare our results with those obtained in laboratory studies. At the outset, we make two qualifications. First, we studied Torpedo californica, the only common species of electric ray off southern California, while most previous research has been conducted on species of Torpedo found in the North Atlantic or in the Mediterranean Sea. Thus, any dissimilarities may result from differences among species rather than between methods. Second, we wanted to assess the natural variation in predatory behavior among different rays. Since we never studied previously-tagged individuals, all of our measurements were statistically independent. Thus, we feel the large variation we observed in behavior and electrical activity is accurate and reflects individual differences in the motivation to feed due to recency of feeding and other ecological and physiological factors. Previous laboratory work on electric rays (e.g. Belbenoit 1981, 1986) described in detail the behavior of certain individual rays and did not provide overall summary statistics. This, coupled with repeated observations on the same individuals, makes it difficult to compare our variation with that observed in the laboratory. Overall, our field measurements of electrical activity agree quite closely with laboratory results. The individual EODs we recorded were monophasic DC pulses averaging --5 ms in duration, similar to previous findings (Bennett et al. 1961; Bennett 1971; Belbenoit and Bauer 1972; Michaelson et al. 1979). Our maximum amplitude measurement of 45 V on a single Torpedo californica in the field was similar to laboratory recordings for adult T. nobiliana (50 to 60 V: Bennett et al. 1961) and T. marmorata (30 to 50 V: Zimmermann and Whittaker 1974; Mellinger et al. 1978, Belbenoit 1979). Coates and Cox (1942) reported that the peak voltage produced by T. occidentalis was 220 V; however, Bennett et al. (1961) argued that this measurement was in error. The lack of a relationship between the amplitude of the EODs and size of the rays is consistent with previous research on other species of Torpedo. The voltage of EODs increases rapidly in embryos and neonates of Torpedo spp. and reaches a plateau at an age of several months. This age corresponds to a length in T. marmorata of = 13 cm (Mellinger et al. 1978, Belbenoit 1979) and a weight in T. ocellata o f = 15 g (Michaelson et al. 1979). Our specimens ranged in total length from 60 to 110 cm and in weight from =4 to 26 kg, well beyond the size (and age) at which the amplitude of EODs of T. californica probably stabilizes. The temporal characteristics of the pulse train, although variable, were also similar to those reported for other species. The general decrease in instantaneous frequency of EODs during the pulse train (Fig. 2) is similar to that observed in all other species of Torpedo studied. The highest instantaneous pulse rate we recorded was 308 EODs s-1. Maximum rates in other Torpedo species range from 100 EODs s -1 in T. nobiliana (Bennett et al. 1961) to over 400 EODs s -1 in T. marmorata (Belbenoit and Bauer 1972). We found that T. californica produced up to 414 EODs during the initial pulse train. The maximum number recorded during predatory attacks in other species was 394 for T. marmorata (Belbenoit and Bauer 1972) and 340 for T. ocellata (Belbenoit and Moller 1971). We note, however, that T. californica can produce many more EODs, especially when fish prey are not ingested after the initial attack or when rays attack a second prey; we recorded a maximum of 1234 EODs produced when a ray attacked two fish (Fig. 2j). Water temperature influences the pulse-train characteristics. The positive relationship between average instantaneous pulse rate and water temperature within the LF treatments corresponded to a Qa0 of 1.75 between 14 and 24 ~ this is slightly below the Qlo of 2 reported by Auger and Fessard (1928). Similarly, the positive trend between maximum pulse rate and water temperature also has been observed in Torpedo marmorata (Belbenoit 1979, 1986). Radii-Weiss and Kova6evi6 (1970) reported that in vitro preparations of electric organ tissue of T. marmorata and T. ocellata displayed decreased amplitudes and increased latencies after stimulation at temperatures below 15 and 10~ respectively. The authors speculated that these rays may not use their electric organs during the autumn and winter when water temperatures are low. The motor patterns of Pacific electric rays were very similar to the "jumping predation" of Torpedo marmorata described by Belbenoit and Bauer (1972) and Belbenoit (1981, 1986). During our daytime tests, rays ambushed prey by quickly jumping over the top of them from the sand. During nighttime tests~ rays actively searched for prey and lunged at prey from in front or below them with vigorous thrusts of their tail. The rays started discharging just when they began the "jump" phase, similar to T. marmorata (Belbenoit 1981, 1986). The experiment with four prey treatments was designed to assess the importance of various stimuli between initiation of an attack and ingestion (Table 1). The three treatments involving fish (live, freshly-speared, or frozen) all triggered a high percentage of attacks. The lack of a clear difference in latency and any of the pulse-train characteristics among these three treatments (Fig. 3 b - f ) , however, does not enable us to eliminate visual, mechanical, electrical or chemical stimuli in prey detection by electric rays (Table 1). Belbenoit (1981) observed that Torpedo marmorata attacked fast-moving inanimate objects as well as fish and concluded that mechanical stimuli were necessary. Our observation on the ray that attacked two fish (Fig. 2j) supports this hypothesis: although a subdued fish was closer, the ray turned and initially attacked a rapidly-swimming fish twice as far away. The SBF provoked far fewer attacks (Fig. 3 a). The major stimuli associated with the SBF trials were the electrical field and the slow and steady movements and visual cue when the apparatus was positioned in front of the ray. Belbenoit (1986) showed that T. marmorata attacks plastic poles (without electrodes) moved near the anterior margin of their disk, presumably due to the mechanical and possibly visual stimuli. It is unlikely that the attacks in our experiments with the SBF were 168 due solely to such stimuli, however, because rays never attacked the paired, unenergized dipole; the probability of all four attacks occurring on the energized d ~ o l e (vs the unenergized dipole) by chance alone is 0 . 5 ~ o r 0.0625. Thus, the SBF treatment provides the first direct evidence that electric rays include electrical cues in their detection and/or assessment of prey. Note, however, that the SBF apparatus consisted of a dipole DC field of fixed amplitude and geometry and lacked modulation that is characteristic of live fish (Kalmijn 1978). This, and the lack of "natural" mechanical, visual, or chemical stimuli, may explain the low proportion of attacks (Fig. 3 a) and long latency (Fig. 3 b) with the SBF apparatus. Our finding that LF and FSF treatments resulted in high percentages of ingestion compared to the FF (Fig. 3 a) indicates that electric rays rely on additional feedback from their prey after the initial attack. The major differences in stimuli between FF and the other two fish treatments were the lack of an appreciable bioelectric field and the absence of muscle twitches induced by the EODs (Table 1); vision during this stage cannot play an important role because the prey is beneath the disk, out of view of the dorsally-located eyes. Belbenoit (1981) concluded that mechanical reception, which probably involves tactile stimulation of the free nerve endings in the skin and acoustic stimulation of the lateral line, appears to provide information on the location of the prey underneath the disk. Our observations support this hypothesis. Rays consistently enveloped prey in their disk during daytime attacks from the sand and nighttime attacks in midwater. In addition to concentrating the electric field over the prey, this envelopment orients mechanoreceptors around the prey, probably providing three-dimensional mechanical information on the prey's activity and location. This mechanoreception may be augmented by electroreception via the single row of ampullary pores that encircles the outer edge of each electric organ. The response of prey to EODs further contributes to the mechanical stimuli. Fish are often tetanized and quiver after the initial attack (Bray and Hixon 1978 and present authors' personal observations), which probably provides constant mechanical stimuli. Additionally, rays often produced EODs after the initial attack (Fig. 2). During the FSF treatments, we often felt the freshly-speared fish twitch through our spear while it was trapped beneath the disk, probably as a result of these sporadic EODs. The muscle twitches induced by these sporadic EODs may enhance mechanical location of prey beneath the disk. Rays responded faster to LF and FSF presented at night than during the day. Bray and Hixon (1978) noted that rays were more active at night and probably foraging at that time; therefore, it was not surprising that rays attacked more readily at night. Slower daytime responses might be attributable to the apparent resting state of the rays. Because of this low daytime-activity state, rays may require additional time to become aware of potential prey. More information is needed on how rays detect prey at night. To date, including this study, most observations have been made on rays resting on sand during the day. In this ambushing mode, the ray and surrounding substrata are immobile when prey move nearby, and the prey is always at the same level or above the ray. At night, on the other hand, rays swim slowly or drift in pursuit of fish prey (Bray and Hixon 1978), prey may or may not be mobile (depending on their nocturnal behavior), and prey is often beneath the moving ray, out of sight. These nocturnal differences in predator and prey behavior create a different suite of stimuli than those during the day and can provide a challenging but rewarding area of future research. Acknowledgements We thank D. and D. Neilson, R. Watkins, and especially M. Ezcurra for their help in the field. We thank G. Lowe for her review of this manuscript and her aid in the field. Financial support was provided by the National Geographic Society, Committee of Research and Exploration through Grant # 3881-88, and the Scholarly and Creative Activity Committee at California State University, Long Beach. References Auger D, Fessard A (1928) Quelques donn6es relatives a la d6charge 61ectrique des torpilles. Bull Stn biol Arcachon 25: 189207 Belbenoit P (1970) Comportement alimentaire et d6charge 61ectrique associfie chez Torpedo marmorata (Selachii, Pisces). Z vergl Physiol 67:205-216 Belbenoit P (1979) Electric organ discharge of Torpedo (Pisces); basic pattern and ontogenetic changes. J Physiol, Paris 75: 435441 Belbenoit P (1981) Sensory control of predatory and defensive behaviours in Torpedo marmorata (Pisces). In: Szab6 T, Czdh G (eds) Sensory physiology of aquatic lower vertebrates. Pergamon Press, Akad6miai Kiad6, Budapest, p. 235-254 Belbenoit P (1986) Fine analysis of predatory and defensive motor events in Torpedo marmorata (Pisces). J exp Biol 121:197-226 Belbenoit P, Bauer R (1972) Video recordings of prey capture behaviour and associated electric organ discharge of Torpedo marmorata (Chondrichthyes). Mar Biol 17:93-99 Belbenoit P, Moiler P (1971) Relations entre la taille de la torpille (Torpedo marmorata) et les charact6ristiques de sa ddcharge alimentaire. J Physiol Paris 62 (Suppl l): 121-122 Bennett MVL (1971) Electric organs. In: Hoar WS, Randall DJ (eds) Fish physiology. Academic Press, New York, p. 347-491 Bennett MVL, Wurzel M, Grundfest H (1961) The electrophysiology of electric organs of marine electric fishes. I. Properties of electroplaques of Torpedo nobiliana. J gen Physiol 44: 757-804 Bray RN, Hixon MA (1978) Night shocker: predatory behavior of the Pacific electric ray (Torpedo californica). Science, NY 200: 333-334 Coates CW, Cox RT (1942) Observations on the electric discharge of Torpedo occidentalis. Zoologica, NY 27:25-28 Crowley PH (1992) Resampling methods for computation-intensive data analysis in ecology and evolution. A Rev Ecol Syst 23: 405 - 447 Kalmijn AJ (1972) Bioelectrie fields in sea water and the function of the ampullae of Lorenzini in elasmobranch fish. University of California, Scripps Institution of Oceanography. La Jolla, California (SIO Ref 72-83) Kalmijn AJ (1978) Electric and magnetic sensory world of sharks, skates, and rays. In: Hodgson ES, Mathewson RF (eds) Sensory biology of sharks, skates, and rays. US Government Printing Ofrice, Washington, DC, p. 507-528 Mellinger J, Belbenoit P, Ravaille M, Szabo T (1978) Electric organ development in Torpedo marmorata, Chondrichthyes. Devl Biol 67: 167-188 169 Michaelson DM, Sternberg D, Fishelson L (1979) Observations on feeding, growth and electric discharge of newborn Torpedo ocellata (Chondrichthyes, Batoidei). J Fish Biol 15: 159-163 Pfeiffer W (1961) Zur Biologie des Zitterrochens (Torpedo marmorata Risso). Pubbl Staz zool Napoli (I: Mar Ecol) 32: 167-171 Radii-Weiss T, Kova6evig N (1970) Influence of low temperature on the discharge mechanism of the electric fish Torpedo marmorata and T. ocellata. Mar Biol 5:18-21 Sokal RR, Rohlf FJ (1981) Biometry. The principles and practice of statistics in biological research. 2nd ed. W. H. Freeman & Co., New York Zimmermann H, Whittaker VP (1974) Effect of electrical stimulation on the yield and composition of synaptic vesicles from the cholinergic synapses of the electric organ of Torpedo: a combined biochemical, electrophysiological and morphological study. J Neurochem 22:435 -450