Quantal Duration of Auditory Memories

6. H. Van de Velde, I von Hoegen, W Luo, J. R. Parnes, K. Thelemans, Nature 351, 662 (199I), S Muthukkumar and S. Bondada, lnt lmmunol. 7, 305 (1995). 7. A. C Lankester, G M. W, van Schijndel,J. L. Cordel, C J. M van Noessel, R. A W, van Lier, Eur. J. lmmunol. 24, 812 (1994); C. Jam~n,P M Lydyard, R. Le Corre, P Y. Younou, Scand. J. lmmunol. 43, 73 ( I 996). 8. A. Tarakhovsky, W Muler, K Rajewsky, Eur. J. lmmunol. 24, 1678 (1994). 9. R. J Noele et a/., Proc. Natl. Acad. Sci. U.SA. 89, 6550 (1992): D C. Parker, Annu. Rev. Immunol. 11, 331 (1993):A W. Heath, W W. WU, M. Howard, Eur. J. Immunol 24, 1828 (1994). 10 S. Reid, R. Cross, E C. Snow, J, Immunol. Meth 192, 43 (1996); L McEvoy, R. A Wl~amson,B. J. Del Buono, J. Leuk Biol. 44, 337 (1988); V. A, e r a , C E Perandones, L. L. Stunz, D A. Mower, R. F. Ashman, ;i,'lmmunol. 151, 2965 (1993), F. Beloc, F Lacombe, P. Bernard, D Dachary, M. R Boisseau, Cytometry 9, 19 (1988). 11. G. Bikah etal., unpubl~shedresults. G L. Warner, D. W Scott, lnt. Immunol 4, 12 L. 6 . LIOU, 15 (1992). 13 A. Tarakhovsky etal., Science 269, 535 ( I 995) 14 J. W. Rooney, P. M Dubois, C H S~bley,Eur. J, lmmunol. 21, 2993 (1991) 15 J LIU,T C Chles, R. Sen, T L Rothsten, J, lmmuno1 146, 1685 (1991). 16. L M. Dale and T. L. Rothsten, J. Exp. Med. 177, 857 (1993) 17. M. Diaz-Meco etal., EMBO J. 13, 2842 (1994). 18. S. H Shen, L. Bastien, B. I. Posner, P. Chretien, Nature 352, 736 (1991), R J Matthews, D. B Bowne, E Flores, M. L. Thomas, Mol. Cell Biol. 12, 2396 (1992), J. Plutzky, 6 . G Neel, R. D Rosenberg, Proc Natl. Acad. Sci U.S A. 89, 1123 (1992), T. Yi, J L Cleveland, J N. hle, Mol Cell. Biol. 12, 836 ( I 992). 19. D R. Plas et a/., Science 272, 117 (1996); G. M Doody et a1 , ibid. 269, 242 (1995). 20. Control mce were 129. The mutant CD5-I- mouse strain was der~vedby intercrossng and subsequent Inbreeding of heterozygous l~ttermates born to C57BU6 females, mated with the male chimeras, obtaned by injectng CD5 gene targeted embryonic stem (ES)cells nto C57BU6 blastocysts (8).Splenic and peritoneal washout cells were obta~nedfrom 7to 12-week old mlce, and depleted of T cells by treatment with a cocktall of ant~bodes(anti-Thy 1.2, ant-CD8, and anti-CD4) followed by rabb~tcompement. Splenic and peritoneal B cells were depleted of macrophages by plastc adherence overn~ghtat 37°C in 5% GO,. B-1 cells were further purlfled by dual staning with Mac-I and B220and soring of the double pos~t~ve cells with a FACStar (Becton D~ck~nson) flow cytometer. Cells (2.5 x lo5) in 0 2 ml of lscove F-12 med~a, 5% fetal calf serum were cultured In flat bottom m~crotiterwells. 21. P. S Rabinovich, C H, June, A Grossman, L. A. Ledbetter, J. Immunol. 137, 952 (1986), N Muthusamy, A R. Baluyut, B. Subbarao, ibid. 147, 2483 (1991). 22. Peritonea B cells were pelleted and resuspended In 100 pl of Ho342 (5 p.g/m) (MolecularProbes) After a 30 m n lncubaton at 37°C In the dark, the cells were peleted and resuspended in 100 pI of a 0 4 pglml stock solutlon of MC540 (Molecular Probes) After a 20-min incubation at room temperature ( ~ the n dark), 1 p g of antl-B220 conjugated to fluoresce~n~sothlocyanate (FITC) (Sigma) was added, and incubated on Ice for 30 m n . The cells were then peleted, resuspended in 1 ml of PBS and 10,000 cells were analyzed immedately on a FACStar flow cytometer (Becton Dickinson).The gates to distinguish MC540 dull (RI, R2) and br~ghtcells (R3, R4) were set by looking for a natural break In the staining profle of the resting or cycl~ngB cells. The CD5-/- m c e required different gates, presumably due to h~gher background stanng of the untreated cells with MC540, which occurs whenever the responding populat~onis actvated (70).Data with per~tonealB cells from widtype 129 mice were shown In F I 2~ and simlar results were obtained with per~toneal B cells from BALB/c mlce 23. Control and CD5-/- 6-1 cells were pur~fiedas described (20) Cells (2 x lo5) were Incubated w ~ t h F(ab'), GamlgM (Cappel)or normal goat IgG (Sigma) for 60 min at 37°C in 5% CO, The cells were transferred onto mlcroscope s d e s by cytocentrfugation, f~xedw ~ t h4% paraformaldehyde, and permeablzed in 0.05% Triton X-100 (60-Rad) Cells were subsequently incubated wlth anti-NF-KB p65 (Santa Cruz Biotech) for 2 hours, then with botnylated anti-rabb t - g G (Vector) for 1 hour. Cells were finally ncubated with avidin-FITC (Vector) for 30 mln, and examned on a confocal laser scannlng mlcroscope (Molecular Dynamcs, Sarastro 2000). To allow for quant~tat~ve comparisons of the relative fluorescence between the cells, the lntensty of the laser beam and the sens~t~vity of the photodetector were held constant. The "Imagespace" software suppl~edby the manufacturer was used to obtan the values for average ntens~tyof nuclear staining Several f~eldsof cells were examned on each slide 24. 6-1 cells (lo7)from w~d-typeor CD5-/- m c e were stimulated w ~ t hF(ab'), GamlgM or goat-lgG for 1 hour at 37°C In 5% CO,. Nuclear proteins were obta~nedas described (25) Protein concentrat~ons were standardzed (BCA, Pierce), and the electrophoretc mobilty sh~ftassay was done w t h a NF-KB bindng proteln assay kit (G~bco-BRL),folow~ngthe procedure of the manufacturer 25 C L Dent and D. S Latchman, in Transcription Factors, D. S Latchman, Ed (Oxford Un~v Press, New York, 1993), pp. 1-26 26. All an~malprocedures were conducted ~n accordance w ~ t hthe Unversty of Kentucky "Guide for the Care and Use of Laboratory Animals 27 We thank A. Kapan and 6 . T Spear for critical review of the manuscript, R. Hardy for a breedng pair of the CD5-/- mlce, M. Howard for monoclonal antbody to CD40, M. Mattson for the use of the confocal laser mlcroscope, J. Strange and R. Cross for flow cytometry, and R S. Matt~nglyfor technical help Supported by NIH grants A121490 and AGO5731 to S.6 " 14 August 1996, accepted 4 November 1996 Quantal Duration of Auditory Memories Sek Jin Chew,*? David S. Vicario,?Fernando Nottebohm?$ Neuronal responses in the caudomedial neostriatum (NCM) of adult zebra finches (Taeniopygia guttata) decreased upon repeated, unreinforced presentations of conspecific song, calls, or other complex sounds. This "stimulus-specific habituation" is a form of learning, and its spontaneous loss, a form of "forgetting." Spontaneous forgetting occurred only at narrowly defined times (2 to 3, 6 to 7, 14 to 15, 17 to 18.5, 46 to 48, or 85 to 89 hours after first exposure to a stimulus), determined by stimulus class, number of presentations, and interval between presentations. The first five forgetting times coincided with periods when gene expression and protein synthesis in NCM were required for maintenance of the longer lasting (85 to 89 hours) habituation. The number of successive episodes of gene expression induced by a stimulus, but occurring long after stimulus presentation, appears to determine the quanta1 duration of auditory memories. T h e songs and calls of songbirds have characterlstlcs that can be used for species identif~cationand individual recognition (1, 2). In previous experiments, we used multi-unit activity (MUA) data to show (I) that auditory responses in populations of neurons In the zebra flnch NCM habituate specifically to individual song stimuli; (it) that this habituat~oncan be long-lasting; and (iii) that the duration of habituation is longer for conspecific songs than for white noise, pure tones, or some exemplars of heterospecific sounds (3). We showed, too, that this habituation was anatomically circumscribed: It occurred in caudal, but not in rostral, NCM 13). stations of , , NCM is one of the h~ehest " the ascending auditory pathway (4). Here we describe habituation at the single-unit level Laboratory of An~malBehav~or,Rockefeller Un~versity, 1230 York Avenue, New York, NY 10021, USA *Present address: S~ngaporeNat~onalEye Center, 11 Third Hospital Avenue, Singapore 031 6, S~ngapore. ?All three authors contributed equally to this work. :l:To whom correspondence should be addressed at Rockefeller Univers~ty Field Research Center, Tyrrel Road, Mlbrook, New York, NY 12545, USA. SCIENCE VOL. 274 13 DECEMBER 1996 and its relation to MUA habituation. We then use recordings of MUA to determine in a svste~naticmanner how st~mulusclass, interstimulus interval (ISI), number of stimulus presentations, and manner of presentation affect the durat~onof stimulus-specif~c habituation in neurons of caudal NCM. FInally, we examlne the relation between the duration of habituation and RNA and proteln synthesis. The fir~ngrates of, angle neurons in caudal NCM were initially high and then decreased upon repeated presentations of the same conspecific song (Fig. 1) (5).After 100 presentations there was little if any further decrement in resoonslveness. The reduction in firing rate was not selectlve for any one subset of the song's components but affected the whole song (Fig. 1C).The stimulus-saecific habituation seen in multiunit recordings does not appear to result from individual neurons "tuning in" to particular features of a sti~nulusand ceasing to resoond to the rest of that stimulus, but rather from a general reduction in the responsiveness elicited by that stimulus. 1909 -2 4 .-- 60 2 A , d 50 ~ -2 40 A e .3 30 ; ,d 20 E .-, L 10 ii 0 0 0.5 1.O 1.5 2'5 2.0 Time (s) Single units exposed t o 50 presentations o f a song gave 21 t 1 (mean 5 SE) spikes per second wheii that song had n o t been heard before ("novel," 66 units), 12 1 spikes per second when that song had been heard 6 t o 40 hours earlier ("remembered," 99 units), and 21 2 spikes per second when that song had been heard 50 t o 100 hours earlier ("forgotten," 51 units) (6). T h e firing rates elicited bv novel and forgotten songs were similar and significantly higher than those for remembered songs (unpaired t tests, P < 0.0001, two-tailed). These observations o n stimulus-soecific habituation at the single-unit level closely parallel observations reported earlier for MUA in NCM (3). + + - 10 20 30 40 50 60 70 Iteration number w , 10 2b , 30 40 50 60 70 80 90 100 Iteration number W e n e x t investigated the factors that determine multi-unit habituation (7). In our standard MUA experiment, a bird is exposed to playbacks of conspecific song at two times. During an initial training stage, birds (n = 49) heard 200 iterations o f the same novel song. T h e n hours or days later, the same song was presented 100 timesthe "testing" stage. T h e IS1 was 11 s, during training and testing. MUA responses were recorded in caudal NCM w i t h tungsten m i croelectrodes (3). T h e decrease in responsiveness during the first 100 iterations o f a novel or familiar stimulus was used to calculate the habituation rate (8) (Figs. l and 2A). N o v e l songs-that is, songs w i t h w h i c h a bird had n o t been trained earlier- 0.00 " "so 0 " 0 OD 0 0 0 10 20 30 40 50 60 70 80 90 Time from training to testing (hours) Fig. 2. (A) Examples of MUA responses to the first 100 presentations of a novel (solid symbols) and a familiar (open symbols) conspecific song at a single site in caudal NCM. MUA response amplitudeswere normalized to the response on the flrst presentation of the novel or familiar song, respectively; the first response to the familiar song overlies that for the novel one. The bird had been trained wlth the familiar song 6 hours earlier. The habituationrates for the novel and familiar songs in this example were 0.42 and 0.16, respectively, and the corresponding best fit lines are plotted for each set of polnts (8). (B)Time course of spontaneous loss of habituation for conspecific song (200 iterations at an IS1 of 11 s). Habituation rates (8)were measured at various training-to-testing intervals in 23 birds, with a total of 10 to 16 different songs per bird, The rates for novel songs are shown at time zero. The time from the onset of training until the time at which the habituation rates became not significantly different from the rates for novel song was used to define the duration of habituation (10). 1910 SCIENCE VOL. 274 90 100 0 Fig. 1. Comparison of single-unit and multi-unit activity in caudal NCM in response to 100 iterations of a novel song. (A) Amplitude waveform of the conspecific song stimulus, (6)Sound spectrogram of the song in (A), (C) Raster of activity In a single unit (5); each dot represents the time of occurrence of an action potential,and each row represents one song presentation (sequence ordered from bottom to top), (D)Mean firing rate (spikes per second ? SE) (5)of the response elicited in the slngle unit shown in (C) over each group of 10 trials (f~lled circles) and normalized MUA amplitude (8) for each trial recorded at the same site (open circles).The MUA habituation rate at this site was 0.38 (Fig:3) (8). induced a h i g h habituation rate (range: 0.30 t o 0.50); familiar songs induced low habituation rates (ranee: 0.01 to 0.25). In addition, the amplitudve o f the response to the first presentation was 14% lower for remembered than for novel songs when these recordings were made at the same site (time between training and testing w i t h the 0 a 80 BER 1996 Time (hours) Fig. 3. The protocol for studylng the tlme course of habituation rates in NCM In a slngle blrd. Traming consisted of presenting sequentially 200 iterations of each of up to 16 novel stlmuli (A to P) at an IS1 of 11 s (total time for each stimulus = 36.7 min). Testing consisted of recording MUA during 100 presentations of each of these now familiar stimuli at an IS1 of 11 s (which remained constant even if the IS1 used in training was, as in some experiments, shorter or longer). In experiments that assessed spontaneous forgetting, stimuli were tested in the reverse order (P to A) to create a wide range of training-to-testing intervals during a single recording session in each bird (13). In those experiments that tested for the importance of RNA and protein synthesis in memory retention, a single dose of CYC or ACT was injected (arrow) into NCM at various delays after onset of training. Testing began 1 hour after injection, starting with the first song presented during training (A to P). This protocol yielded 16 different training-to-injection intervals for each bird tested. familiar song, 6 to 8 hours; n = 42 paired comparisons; paired t test, P < 0.002). T o examine habituation rates at many different intervals after training, we sequentially played multiple stimuli to each bird during training; then, during testing, we played the same stimuli again in reverse order (Fig. 3). Habituation rates to each familiar song remained significantly lower (9) than those elicited by novel songs until -47 to 48 hours after onset of training, when there was a relatively abrupt transition to the higher habituation rates elicited by novel songs (Fig. 2B) (10). We defined duration of habituation as the time from initial exposure to a stimulus during training to the time when the habituation rate for that stimulus became statistically indistinguishable from the rates for novel songs. The duration of habituation was affected by the type of sound presented (I I); it was the same for all familiar exemplars of a given stimulus class and differed among stimulus classes (Fig. 4A). Human speech and canary (Serinus canaria) song produced memories lasting 3 hours; songs of the Bengalese finch (Lonchura striata), a member of the same Estrildid family as the zebra finch, 6.5 hours; all reversed conspecific vocalizations, 6.5 to 7 hours; conspecific male long calls (2, 12), 18.5 hours; and conspecific male song and female long calls (2, 12), 47 to 48 hours. These memory durations were the same whether our recording electrode was in the right or left NCM (13). Although gender features of the stimulus determined the memory duration for male and female calls, there were no differences in memory duration between male and female subjects for any type of stimulus (13). The effectiveness of conspecific song in inducing long-lasting stimulus-specific habituation was affected by the IS1 used during training (Fig. 4B). At an IS1 of 3.5 s, immediate habituation to novel songs was weak (14). These "familiar" songs were subsequently regarded as novel even when tested 1 hour later. The duration of habituation for ISIs of 3.62 to 3.87 s was 6.5 to 7.5 hours; for ISIs of 4.0 to 4.5 s, 13.5 to 14 hours; for ISIs of 4.75 to 8 s, 17 to 18 hours; for ISIs of 9 to 30 s, 48 hours; and for ISIs of 35 to 54 s, 87 to 89 hours. Over this range of training ISIs (testing IS1 held constant at 11 s), with corresponding increases in training time, changes in memory duration occurred in abrupt steps; there appeared to be different thresholds that the IS1 had to exceed before the next quantum in memory duration occurred. We also studied the effect of the number of stimulus iterations on the duration of habituation. As the number of iterations of conspecific song increased gradually during training from 30 to 1000, at a constant IS1 of 11 s (training time ranged from 5.5 min to 3 hours), we again saw stepwise increases in memory duration (Fig. 4C). No lasting habituation was produced when only 30 to 40 iterations were presented. Habituation lasted 6.5 to 7 hours with 50 to 140 iterations, 17.5 to 18.5 hours with 150 to 185 iterations, and 47.5 hours with 185 to 200 O.OO1 0 2 4 iterations. There was no further increase in the duration of habituation when birds were trained with up to 1000 iterations. One thousand iterations at an IS1 of 11 s required 3 hours for training, which is as long as it took to present 200 iterations at an IS1 of 54 s, yet the habituation lasted 47.5 and 87 hours, respectively. Total training time 6 ' 8 1 0 1 2 14.1618'26'i2'i4 % "40. ''60' "so' ' i& Time from training to testing (hours) Fig. 4. (A) Duration of long-term habituation differed among stimulus classes. The habituation rates (TSE)are plotted as a function of the training-to-testing interval. The loss of long-term habituation to these familiar stimuli occurred at distinct times characteristic of the stimulus class, with all exemplars of that class being forgotten at the same time. The longest memory durations are seen for conspecific songs and female calls. In the following sentence,the number after each stimulus class indicates the number of exemplars tested for that class: HUM (wordsfrom human speech),7; CAN (canarysong),5; BEN (Bengalesefinch song),3; MC (malezebra finch long call),6; FC (femalezebra finch long call),6; SG (zebrafinch song),16; reversed conspecific vocalizations: RMC, 6; RFC, 6; RSG, 16. Forty-nine birds were tested with these stimuli. (6)Duration of long-term habituation depended on the ISI. IS1 was systematically varied during training from 3.5 to 54 s, with 200 sequential iterations of each conspecific song stimulus in a total of 58 birds; training time for each song ranged from 11.6 min to 3 hours. Testing was always done at an IS1 of 11 s. Small increments in IS1 produced large, step-like increases in the duration of long-term habituation. (C). Duration of long-term habituation depended on the number of stimulus iterations.The number of sequential presentations of each conspecificsong was systematically varied during training from 30 to 1000, with an IS1 of 11 s in 52 birds. Small increases in the number of iterations produced large, step-like increases in the duration of long-term habituation. In all curves, at least four birds are represented per time point within 2 hours (graphson left) or 5 hours (graphson right) on either side of a step change leading from remembered to forgotten. SCIENCE VOL. 274 13 DECEMBER 1996 did not, by itself, determine the duration of during which injection of CYC or ACT human speech, canary song, and male and blocked long-term habituation by varying female zebra finch long calls, when habituthe ensuing habituation. Spaced training has been shown to pro- the time ela~sedbetween trainine and in- ation to anv of these stimuli lasted lone duce a longer behavioral memory than jection while using the same intervals be- enough ( ~ i ~ . ' to 4 encompass ~) any of thesi massed training in Drosophifu (15). We ex- tween injection and testing (Fig. 3). ~eriods. After CYC injections (protocols 1 and Each of the characteristic times for the amined whether the same held true for neuronal memory in NCM by comparing 2) (Fig. 5), the habituation rates for familiar spontaneous forgetting of a familiar sound the duration of habituation in a paradigm songs were similar to those induced by nov- corresponded closely to one of the sensitive where iterations of each stimulus were ei- el songs when injections occurred 0.5 to periods when blockage of RNA or protein ther presented in a single group (massed) or 3.0, 6.5 to 7.0, 14.0 to 15.0, 17.5 to 18.5, synthesis resulted in a loss of habituation in several smaller groups separated in time 33.0 to 38.0, or 44.0 to 48.6 hours after (Fig. 6). However, none of the stimulus types (spaced) in a balanced design (16). When onset of training with the song tested (t = or training paradigms used produced spontawe compared these two paradigms, using -0.11 to -1.47, P > 0.05). At these times, neous forgetting that corresponded to the 200 iterations for each sow. habituation the rates were also significantly different 33- to 38-hour time window durine which was lost by 48 hours for the sings presented from those obtained for the same songs on RNA synthesis and protein synthisis was in the massed-training manner, as expected, the saline-injected side (t = -2.34 to reauired for habituation to be maintained. but spaced training produced habituation -8.45, P = 0.0063 to 0.035). After ACT Those are the facts that must be.evaluthat persisted for 89 hours (17). When we injections (protocols 1 and 2) (Fig. 5), the ated. The term "habituation" has been used used a spaced-training paradigm (18) to habituation rates to familiar songs were sim- in the past mainly to refer to a situation in present canary song and human speech, ilar to those induced by novel songs when which an animal ceases to give behavioral these heterospecific stimuli were regarded injections occurred 0.5 to 1.5, 6.0 to 6.5, respo%es to repeated presentations of an as novel when testing started at 4 to 5 hours 14.0 to 14.5, 17.5 to 18.5,32.5 to 36.5, and unreinforced stimulus (22). The same term (training took 3 hours), suggesting that in 44.0 to 47.0 hours after onset of training has also been used to denote a decrement in this instance the spaced training paradigm (t = -0.14 to -1.32, P > 0.05). The rates neuronal responsiveness under conditions had added little if any to the duration of at these times were significantly different in which this decrement was known to be habituation. from those for the same songs recorded from related to behavior (23). We do not know We also investbted the mechanism of the control side (t = -3.64 to -8.27, P = whether the single-unit and multiple-unit quantal memory. L l i e r work had shown 0.0004 to 0.04). habituation that we saw in caudal NCM that when stimulus-s~ecifichabituation of Iniections of CYC or ACT at times oth- were related to chanees in behavior. but it is the auditory responses in neurons of the er &in the six sensitive periods defined apparent that the cianges in neuional rezebra finch NCM was long lasting, its persis- above did not affect the retention of lone- s~onsivenesswere related to ex~erience. Exposure to conspecific vocalizations tence in awake animals could be interrupted term habituation for the familiar stimulus by blocking RNA and protein synthesis in (Fig. 5). We infer that long-term habitua- elicited longer lasting habituation than any NCM during two sensitive periods, 1 to 3 tion that lasted for up to 80 hours after of our other auditory stimuli. A species' own hours and 6 to 7 hours, respectively, after training with conspecific song depended on set of signals may often have the longest onset of exposure to a particular stimulus, RNA and protein synthesis that occurred claim to memory duration because of the but not during the intervening time (3). during the six sensitive periods defined relevance of the information it conveys. If This earlier observation was com~atiblewith above (21). Very siinilar periods of sensitiv- so, conspecific signals have special advanthe possibility that the second G v e of gene ity to CYC and ACT were found for the tages for studying the mechanisms that deexpression was, as suggested also for other retention of habituation to exemplars of termine memory duration. Had we used systems (19), part of a two-step mechaniim of memorv consolidation. However. the quantal nature of the memory durations just described was so robust that we wondered whether long-term habituation consisted of consecutive memory segments, each induced by a specific molecular process. To study the role of RNA and protein synthesis in the maintenance of long-term habituation in NCM, we trained animals with one of three protocols: (1) massed training at an IS1 of 11 s (n = 28), (2) massed training at an IS1 of 40 s (n = 14), or (3) spaced training at an IS1 of 11 s (n = 4) (16). At various intervals after the end of training, but well within the time when the habituation induced would still have been Fig. 5. Sensitiveperiods when injectionsof CYC and ACT into NCM resulted in loss of stimulus-specific present, each bird was injected into the habituation. The habituation rates (3,8)shown (mean 2 SE) were plotted as a function of the interval right or left NCM with either the RNA between the onset of training with a particular song and the time of injection. CYC curve (red):data from synthesis inhibitor actinomycin-D (ACT, 0.5 to 19.8hours were obtained with protocol 1 (IS1= 11 s);data from 20 to 51 hours, with protocol 2 (IS1= 40 s)(seetext).ACT curve (blue):data from 0.5 to 19 hours were obtained with protocols 1 and for protocols 1 to 3) or the protein synthesis 2; data from 20 to 87 hours, with protocol 2. ACT spaced curve (green):data were obtained with inhibitor cycloheximide (CYC, for proto- protocol 3. Each point represents data from 3 to 10 birds, with a mean of four birds per point in each cols 1 and 2) (Fig. 3); saline was injected training protocol. Abbreviations: SAL, saline; CYC, massed training, CYC injection; ACT, massed into the contralateral NCM as a control training, ACT injection; ACT Spaced, spaced training (fivesongs, each played 200 times in four groups (20). We identified the sensitive periods of 50 iterations; IS1 = 11 s),ACT injection. See text for other details. - - SCIENCE VOL. 274 13 DECEMBER 1996 only white noise or tones as probes, we would not have discovered that habituation occurs for various, fixed durations of time determined by stimulus class and manner of presentation. The most counterintuitive result from our experiments is the observation that habituation lasted for fixed periods of time-2 to 3, 6 to 7, 14 to 15, 17 to 18.5, 46 to 48, or 85 to 89 hours after onset of trainingand that even when some stimulus oarameters were altered linearly-for example, IS1 and number of iterations-the resulting duration of habituation did not fall along a continuum. Moreover, the same fixed periods emergkd when we used different classes of sound, changed the ISI, or varied the number of iterations. In all instances, the duration of habituation increased by fixed, quantal amounts. The times for forgetting corresponded closely to the sensitive periods during which RNA or protein synthesis was required for maintenance of long-lasting habituation (Fig. 6). Previous studies have reoorted that ACT blocks long-term memory (24, 25) through interference with geSimilarlv. netic transcri~ti'on (23). , , , . CYC blocks protein synthesis necessary for learning-an effect that is reversible (15, 20, 25, 26). In both instances, effectiveness depends on the time relation between training and drug exposure. By extending the interpretation of these earlier studies to our - - present results, we infer that the maintenance of lone-term habituation to a familiar sound required multiple episodes of gene expression and protein synthesis in NCM. Moreover, we suggest that the duration of habituation was determined by the number of successive sensitive periods during which mnemogenic RNA and protein synthesis occurred. If the times for the prolongation of habituation are fixed, then a reference "time zero" must be set. Because the times for forgetting and for sensitivitv to CYC and A?T cirresponded so weli across a diversity of training protocols (Figs. 5 and 6), although some of these protocols took much longer than others (the duration of training with any one stimulus ranged from 36 min to 2.4 hours), we suggest that time zero was set during the first half-hour of stimulation (the level of resolution provided by our methods), and possibly even during the first few presentations of a stimulus. A n extraordinary implication of this hypothesis is that each new stimulus starts its own molecular clock, and that therefore a very large number of such clocks must be running all the time in the brains of awake animals immersed in a sea of sensory stimulation. It is tempting to explain our observations on memory duration by a specific mechanism, but the best we can do is to suggest features that such a mechanism u should have: (i) It must be stimulus-specific. (ii) , , It must have a time zero from which subsequent durations are determined. (iii) It must be able to encode memorv duration in fixed quanta, such that memories can last 3, 7, or 14 hours or longer without expressing durations that fall in-between. ( A mechanism that activates each quantal duration when the input reaches a threshold comes to mind.) (iv) It must be able to integrate input over varying periods of stimulation, so that repeated instances of a particular stimulus can be added to an ongoing record, even when they occur at varying intervals and intermingled with other stimuli. iv) , . It must be able to induce segments of memory duration that occur always in the same order, such that an initial duration of 3 hours is followed by another one of 3 to 4 hours, which in turn is followed by one of 6 to 7 hours, and so forth. (vi) There must be a dependence among successive memory segments, because blockage of RNA and protein synthesis during the sensitive period that initiates a segment causes a memory loss that oersists into at least what would have been the next memory segment (27). The neuronal habituation that we have studied is a form of learning that occurs in the absence of reinforcement (22, 28). Its properties may differ in important ways from other kinds of memory, for example, associative learning. Yet even associative learning relies on stimulus recognition, and SO the differences mav be minor. Our observations raise three hypotheses that may apply to learning in general: (i) memory comes in quantal durations; (ii) these quantal durations are determined bv successive episodes of RNA and protei; synthesis, durations resulting from the with longer " sequential action of several such episodes; and (iii) the duration of long-term memory is determined by mechanisms that are an integral part of learning. - u REFERENCES AND NOTES Stimulus class IS1 (s) Number of iterations Fig. 6. Temporal correspondence between t~meswhen spontaneous forgetting occurred (Fig. 4) and times when RNA and prote~nsynthesis were required for maintaining long-term habituation in different stimulation paradigms (Fig.5).Duration of habituation is plotted for different classes of stimuli (left graph), all numeric scales are logarithmic. In each type different lSls (middle),and numbers of iterations (r~ght); of experiment,spontaneous forgetting occurred only at three to five f~xedtimes. The dashed horizontal lines Indicate the ends of per~odswhen macromolecular synthes~swas required for habituation to be maintained, as determined by injecting CYC and ACT. Abbreviations are as in Fig. 4A. All stimulus classes were trained w~th200 repetitions at an IS1 of 11 s, with the exception of SG40 (conspecificsong, IS1 = 40 s) and SSG (spaced train~ngwith four groups of 50 iterations of conspecific songs). SCIENCE VOL. 274 13 DECEMBER 1996 1. D. E. Kroodsma and E. H. Miller,Acoustic Communication in Birds (Academic Press, New York, 1982). 2. R. Zann, An~m.Behav. 40,811 (1990) 3. S.J. Chew, C.V. M e o , F. Nottebohm, E. D. Jan~is, D. S.Vicario, Proc. Natl. Acad. Sci. U.S.A. 92, 3406 (1995). 4. G. E. Vates, B. M. Broome, C. V. Mello, F. Nottebohm, J. Comp. Neurol. 366, 613 (1996). 5. Zebra finches (97 male and 96female) obtalned from our breeding colony were prepared for long-term recordlng as descrbed (3).Auditory stimuli (11) were played from a speaker placed 0.5 m from the bird in a soundproof experimental chamber (3, 12). During traning the bird was freely moving: electrophysiologica recordlng was carrled out n awake, restrained birds. Insulatedtungsten m~croelectrodeswere used to record physiological activlty in the rlght or left caudal NCM at stes that exhbited habltuaton to novel songs (3). Recording sessions lasted -5 hours. All procedures conformed to an anma use protocol approved by the Rockefeller Unversty Animal Care and Use Commttee. The audtory stmuus and m croelectrode data were d g ~ t ~ z eatd 20 kHz. S n g e u n ~ tact~onpotental waveforms were d ~ g t a yd ~ s cr~m~nated and dspayed as rasters (Exper~menters Workbench, Datawave). S n g e - u n t responses were quantf~edby averaging the frlng rate durlng the st~mu u s per~od(plus the ensuing 100 ms) and then subtract~ngthe f r ~ n grate durng the control perod (500 ms precedng st~mulusonset). T h s mean f~rlngrate per presentaton was used to produce averages, for example, of 10 success~vepresentatons (Fig. 1D) or of the frst 50 presentatons 6. B~rdsthat had heard the testng st~mulusear~erhad been exposed to 200 repettons thereof at an IS1 of 11 s; testng also occurred at an IS1 of 11 s. 7. Slngle un~tsare d~ffcultto "hold" w ~ t ha recordlng electrode for per~odsof more than 30 m n Therefore, most of t h ~ sstudy was done w ~ t hMUA data. 8. The multi-un~t response amptude to each stmuus presentat~onwas calculated by subtractng the rootmean-square (rms)value over the 500-ms precedng st~mulusonset from the rms over the per~odfrom st~muusonset to offset p u s 100 ms. The d~fference between the two root-mean-square values measures the net response per u n ~ttm e and corrects for d~fferencesn durat~onbetween s t m u l ~Each . ampitude was then normalzed to the response ampl~tude on the f~rstpresentaton (typ~callythe largest) and plotted as a funct~onof st~mulusteraton. We used the east-squares method to determne the slope of the stra~ghtlhne that best f~ttedeach set of 100 normallzed responses to a st~mulusthat the b r d had or had not heard dur~ngan earler tranng sesslon. The habtuat~onrate was def~nedas the absolute value of the slope of the best-fit line (Fig. 2A). Hab~tuaton rates were Independent of the absolute response level at any gven s~teand so could be used to compare d~fferentsltes, In d~fferentb~rds,recorded at dfferent lnten~alsafter tranng. We have prev~ously shown that the d~str~but~ons of hab~tuatonrates to novel and faml~arsongs dffer s~gnlf~canty (3). F~gure 2A demonstrates, too, that habtuat~onto t h e f a m ~ a r song occurred, typcally, dur~ngthe frst few repettons of the st~mulus,after whlch responses stayed at a lower level; t took longer for this lower level to be reached by responses to a novel song 9. All statst~cacomparisons were done w t h the Student's t test (P < 0.05, two-ta~led). 10. The dstrbut~onof entrles n Fig. 2 8 suggests that, even though there was a reatvely abrupt change In hab~tuatonrate between 46 and 48 hours, a less marked dr~fttoward h~gherhab~tuat~on rates occurred between 15 and 46 hours after onset of trann g . Our strngent cr~ter~on for f o r g e t t n g h a b ~ t u a t ~ o nrates smlar to those induced by presentatons of a novel song-dd not recognize these changes, which may, however, represent an early forgettng stage that deserves further study 11. These sounds, whch provded a set of conspecfic and heterospecfic s t m u i that the b~rdshad not prevously heard, were dgitized at 20 kHz (Sgna, Englneerng Desgn) The songs and words from human speech were 1 2 to 2 0 s long, and calls were 100 to 400 ms long 12 H B S~mpsonand D S V~car~o, J. Neurosci. 10, 1541 (1990) 13 S J Chew, D S V~carlo,F Nottebohm. Proc Natl Acad Sci U S A 93,1950 (1996) 14. S J Chew, D S V~cario.F. Nottebohm, data not shown 15 T Tuly. T Preat. S C Boynton, M D e Veccho, Cell 79 . - , 95 - - 119941 -- , 16 Seventeen birds were tra~nedw ~ t h200 ~teratlons (massed) each of two songs, followed by 50 terat~onsof each of flve other songs, and thls latter spaced protocol was repeated four tmes At an lSl of 11 s, the spaced-tranng part of this protocol took 3 hours. Another two songs were then played w ~ t hthe massed-tra~n~ng paradgm The total durat~onof t h s protocol was 5 5 hours In a second verson of t h s protocol, b~rds(n = 8) were tra~nedw ~ t h50 terat~ons (massed tra~n~ng. I S = 11 s, total training time = 9 2 m n ) of each of two novel songs, this was then followed by 10 teratons of each of a group of five other novel songs, and the latter sequence was repeated act~vltyat the njecton sites Comrecord phys~oog~cal f~vet~mes(spaced tranng: IS1 = 11 s, total tranng parlson of s~multaneousrecordings n the two s~des t m e = 46 mn). We then played another three songs controlled for nonspecf~ceffects or drug dffuson and uslng the massed-tra~n~ng parad~gm Thus, n 1 5 for the part~cularsongs used n any glven exper~ment hours, we exposed the b ~ r dto 50 terat~onsof f~ve The two s~deswere prev~ousyshown not to d~ffern songs presented In the massed-train~ngparad~gm, hab~tuat~on rates (131 Test st~mulwere presented In the and 50 teratlons of flve songs presented n the same order as durng tra~nng,start~ng1 hour after n spaced-tranng parad~gm ject~on 17. When a total of 50 teratons of each song were used, 21 The presence and durat~onof the f~rstsens~tlve hab~tuat~on was lost after only 7 hours for songs per~odwas not established w ~ t hprotocols 2 and 3 presented ~nthe massed-tranng manner (Fig 4C), because tralnng w ~ t ha sngle song w ~ t he~therof but spaced tranng (761 produced habtuaton that these protocols lasted 2 2 and 2 4 hours The secpers~stedfor at least 43 hours (14). ond, t h ~ r d ,and fourth sensltlve per~ods dur~ng 18. Fve zebra fnches that were tra~nedw ~ t hcanary song w h ~ c hlnject~onsof ACT or CYC n NCM blocked presented n a spaced parad~gmheard four groups long-term hab~tuat~on were very s ~ m l a rregardless , , a total of of 50 lteratons for each of f~ves t m u ~for of whether we used protocols 1, 2 or 3 In a d d 200 teratlons for each st~mulus;thls tranng was t ~ o n protocols 2 and 3 revealed a f~fthand s~xth compared w ~ t h200 massed iterat~onsfor a same senstve perod of mnemogenc RNA and proteln canary song. synthess, w h e retanng those that had occurred 19. G. Grecksch and H. Matthes, Pharmacol. B~ochem. earl~er( F I ~5). . Behav 12, 663 (19801; H Matth~es,n Long-Term 22. P Marer and W. J. Hamilton, Mechanisms ofAnima1 Potent~at~on, M. Baudry and J. L. Dav~s,Eds ( M T Behavior (Wley, New York, 19661. Press, Cambridge, MA, 19941, pp. 233-243; F. M 23. R. D. Hawkns and E R. Kandel., Psvchol. Rev. 91. , Freeman, S. P. R Rose, A. B. Schoey, Neurobiol. 375 (1984). Learn. Mem 63, 291 (19951. 24. P V. Nquyen, T. Abel, E. R. Kandel, Science 265, 20 ACT (40 nl, 50 F M ~ or CYC (40 nl, 1 ng,n), dssolved In 1 104 (199.41 sane, was Injected nto 46 adult male or female zebra 25. H. P. Davs and L. R Squ~re,Psychol. Bull. 96, 518 finches lnto a rlght- or left-hem~sphereNCM recordng (1984). s~tethat exhlblted hab~tuat~on to novel songs. These 26. C. H. Baley and E. R Kandel, n The Cognitive Neu~njectonswere made w~thagass mcroplpette (t~p outer rosciences, M. S. Gazzan~ga,Ed (MIT Press, Camd~ameter,20 Fm).As a control, saneveh~cealone was br~dge,MA, 19951, pp 19-36. injected nto the other hemsphere. The effect~vesphere 27. Thls effect was not~ced,for example, when testng of the ACT or CYC njectons, as determ~nedby Immu10 or 24 hours after onset of tranlng ~nb ~ r d sthat nocytochemlstry, was I~mted to asubregon of NCM (3) receved njectlons of blockers at 0.5 to 3 or 14 to 15 Such njectons d d not affect aud~toryresponses or the hours, respect~vely. ~mmedatehabtuaton of NCM neurons to playbacks of 28. E. R. Kandel, J. H Schwartz, T. M. Jessell,Princ~ples a novel song. The phys~olog~cal effect of these RNA and of Neural Science (Elsev~er,New York, 1991). prote~nsynthesis blockers n NCM was reversed ~n<1 hour, as deterrnned by the loss of long-term habtuaton 29 We thank S. Creel, R. Des~mone,C. Gilbert, N. Hentz, M. E. Nottebohm, M. Young, J. Wailman, T. for st~mulpresented 0.5 hour, but not 1 hour, after ~njectlon[S. J Chew, thesls, Rockefeller Un~vers~ty W e s e , and two anonymous revewers for the~rhepful comments on the manuscr~pt,and N. Naqv~for (1966)],allow~ngfor a farly accurate pnpontlng of the help with the s~ngle-untanayss. Supported by P H s tme when gene expresson or proteln synthess IS necessary for the maintenance of long-term habltuatlon. grants MH18343 (F N.) and MH40900 (D S V.) and The side of drug lnjecton was alternated n s~~ccess~ve by the Herbert S~ngerand the Mary Flagler Cary experments to emnate sde-to-s~debases ~nlnjecton Chartable Trust or recordlng tschnque. Startng 1 hour after drug Injection, Insulated tungsten mlcroelectrodes were used to 19 June 1996, accepted 2 October 1996 Ethylene as a Signal Mediating the Wound Response of Tomato Plants P. J. O'Donnell, C. Calvert, R. Atzorn, C. Wasternack, H. M. 0 . Leyser, D. J. Bowles* Plants respond to physical injury, such as that caused by foraging insects, by synthesizing proteins that function in general defense and tissue repair. In tomato plants, one class of wound-responsive genes encodes proteinase inhibitor (pin) proteins shown to block insect feeding. Application of many different factors will induce or inhibit pin gene expression. Ethylene is required in the transduction pathway leading from injury, and ethylene and jasmonates act together to regulate pin gene expression during the wound response. \ T h e wound rebponse o i tomato plants has been studied for some 25 years and represents a model system for the analvs~sof cell signaling pathways in plants ( I ). Prote~nase P J O'Donnell. C Calvert, H. M. O Leyser, D J. Bowles, The Plant Laboratory. Department of Biology, Univers~ty of York, P O Box 373, York YO1 5YW R Atzorn, C Wasternack, nsttut fur Pflanzenbochem~e (IPB), We~nberg3. 06120 Halle (Saale). Germany 'To whom correspondence should be addressed SCIENCE \'OL.274 13 DECEhtBER 1996 ~ n h i h t o r (pin) genes are up-regulated throughout aerial tissues in response to wounding ( 2 ) . Pin genes are also responslre to c o m p o u n d s a p p l i e d experimentally through t h e transpiration stream of excised leaves and the use of this bioassa\- has identliied a range of positive and negarive regulators. Posltlre regulators (elicitors) include ohgogalacturoniile fragments of pectin polysacchar~des( O G A s ) ( 3 ) ,a n 18mer