Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Vol. 139, Supplement The AmericanNaturalist March 1992 SIGNALS, SIGNAL CONDITIONS, AND THE DIRECTION OF EVOLUTION JOHN A. ENDLER Departmentof Biological Sciences, Universityof California,Santa Barbara, California93106 Abstract.-There is a bewilderingdiversityof signals,sensorysystems,and signalingbehavior. A considerationof how these traitsaffecteach other'sevolutionexplains some of thisdiversity. Natural selection favors signals, receptors,and signalingbehavior thatmaximize the received signals relative to background noise and minimizesignal degradation. Propertiesof sensory systemsbias the directionof evolution of the signals thattheyreceive. For example, females may prefermales whose signals theycan perceive more easily, and thiswill lead to the spread of more easily perceived male traits.Environmentalconditionsduringsignal transmissionand detectionalso affectsignalperception.Specificenvironmental conditionswillbias theevolutionary directionof behavior, whichaffectsthe timeand place of signalingas well as microhabitat preferences.Increased specializationof microhabitatsand signalingbehaviormay lead to biased evolution of the sensory systems to work more efficiently. Thus, sensory systems, signals, signalingbehavior, and habitatchoice are evolutionarilycoupled. These suites of traitsshould coevolve in predictabledirections,determinedby environmental biophysics,neurobiology,and the genetics of the suites of traits-hence the term"sensory drive." Because conditionsvary in space and time,diversitywill be generated. At least some of this variation The immensediversityof species is intriguing. in physical appearance is the result of the diversityof visual and other signals used in various formsof communication,includingattractingand courtingpotentialmates, maintainingterritories, holdinggroupstogether,and minimizing predation. The diversityin appearance is rivaled by variationin the times and places and in the designsof thedevices used to receive wherethe signalsare transmitted them. Can any of this variationbe explained or even predicted? Natural selection favors signals, receptors,and signalingbehavior that maximize the received signalrelativeto backgroundnoise and minimizesignaldegradation. Conditionsduringsignaltransmissionand detectioncan affectthe quality and effectivenessof received signalsbecause bothcan alterthe signal's perceived form.Therefore,a signal's effectivenesswilldepend on the signal'sform,receiver design, and behavior thatdeterminesthe environmentalconditionsduringtransmission. As a result,signals,receptors,and behaviorare not suites of evolutionarilyindependenttraits;theyare functionallyrelatedand are thereforelikelyto influencethe evolutionof one another.The directionof the resultingjoint evolution of signals, receptors, and signalingbehavior is affectedby environmental physics, biophysics,and neurobiology.The processes leading to these biases in the directionof evolutioncan be loosely called "sensory drive," which suggests that sensory systemsand sensoryconditions"drive" evolutionin particulardiAm. Nat. 1992. Vol. 139, pp. S125-S153. ? 1992 by The Universityof Chicago. 0003-0147/92/3907-0007$02.00. All rightsreserved. This content downloaded from 152.14.136.96 on Mon, 3 Jun 2013 09:12:05 AM All use subject to JSTOR Terms and Conditions S 126 THE AMERICAN NATURALIST Environmental f conditlons duringsignal Signaling behavlor: lVsibility _ Umingndseason etc. ofcourtship, sibility topredstr influences, evolutionary indicate drive.Arrows processesofsensory FIG. 1.-The essential to Ryan's effects."Theshadedportionis equivalent exceptfortheone labeled"immediate model. (1990)sensoryexploitation rections(Endlerand McLellan 1988;Endler1989).The purposeofthisarticleis and to to exploresome aspects of sensorydrive,to makea few predictions, theprocessof sensory workon thesubject.I willsummarize encouragefurther makepredictions aboutvisual drive,givean exampleforvisualcommunication, forall sensorymodes. systems,and thendiscusssomegeneralimplications THE PROCESSOF SENSORYDRIVE amongsensorysysthemainevolutionary relationships Figure1 summarizes these underwhichtheyare sentand illustrates tems,signals,and theconditions male traits. Similar interacwiththeexampleof sexuallyselected relationships and otherkindsof signals. antipredator, tionsshouldaffectterritorial, how maletraitsare perthe of sensorysystemdetermine The characteristics mate-choice criteria because theyenwill the evolution of affect Senses ceived. well. If only will be male all available that not equally perceived signals sure and are able to be detected of easily perceived or sometypes components signals choice whereas less used as female will be then these other, criteria, clearly, are mate-choice criteria will be used. not Thus, components easilyperceived This content downloaded from 152.14.136.96 on Mon, 3 Jun 2013 09:12:05 AM All use subject to JSTOR Terms and Conditions SIGNALS AND THE DIRECTION OF EVOLUTION S127 predisposed to evolve in the directionsfavored by sensory characteristics.In some cases, thispredispositionmay even lead to preferencesformale traitsthat have not yet evolved (Basolo 1990). Mate-choice criteriadetermine,or at least stronglybias, thedirectionof evolutionof male traits;preferredmale traitsspread traits.In a sense, the signal characteristicsof the at the expense of unpreferred male traitsevolve to "exploit" the signal-receptioncharacteristicsof females; signals that stimulatethe sensory systemmost stronglyhave an advantage over those that result in less stimulation.Ryan (1990) termedthis process (shaded in fig. 1) "sensory exploitation" and has providedmuch evidence for it in the auditorysignals of frogsand other animals (Ryan 1990; Ryan and Rand 1990; Ryan and Keddy-Hector 1992). However, additionalprocesses are also operating (fig. 1). The signal's evolutionis also affectedby predation,a factwell-knownforboth visual and auditorysignals(Endler 1978, 1983;Tuttleand Ryan 1982). In general, the signal evolves as a local balance between the relative strengthsof sexual selection and predation(Endler 1978, 1980, 1983). For example, if predationis relativelystrongerthansexual selection,thencolor patternswill be morecryptic. If predationis relativelyweaker, color patternswill be more conspicuous and elecloser to those predictedfromthe sensory exploitationmodel. If different then theywill reach different balmentsof the signal are perceived differently, ances between the two selective factors.For example, ifthe predatoris particularly sensitive to yellow and not very sensitiveto red, then, even thoughboth yellow and red may be equally brightto females, the balance between sexual selectionand predationescape will favormale color patternsthatemphasize red (Endler 1978, 1991). The environmentis spatiallyand temporallyheterogeneous,and the physical propertiesof the environmentaffectthe rates of attenuationand degradationof the signal (Lythgoe 1979; Wiley and Richards 1982; Halliday and Slater 1983; Ryan 1985, 1988a; Endler 1986, 1990). Most animals do not signal continuously but,rather,transmitat particularseasons, times,and in particularmicrohabitats; thishas the immediateeffectof ensuringa specificand predictableset of environmental conditions duringtransmission(fig. 1). For a species with a particular effectively only at certaintimesand in male signal,the signal can be transmitted certain places. Males that transmitat these times and in these places obtain more mates than males thattransmitat timesand in places where attenuationor degradationis greater.This is likelyto favorincreasingspecializationof signaling behavior,which leads to better(less attenuationand degradation)and more predictable signalingconditions (fig. 1). This argumentapplies to the colors and scents of fruitsand flowersas well as to animal signals; here, sensorydrive will involve both the plant and its dispersers or pollinators.Specialized signaling behaviormay also affectthe generalmicrohabitatchoice of the species, particularlyin animals. If a species spends much of its time in a particularmicrohabitatand signals only undercertainconditions,thenthe sensorysystemcan evolve to best match theconditions.In fact,thereis muchevidence thatsensorysystemshave evolved to be "tuned" to match the average characteristicsof the environment.For This content downloaded from 152.14.136.96 on Mon, 3 Jun 2013 09:12:05 AM All use subject to JSTOR Terms and Conditions S128 THE AMERICAN NATURALIST example, the spectral sensitivityof manyunrelatedfisheshas convergedto patclasses of water color and ambientlightconditionsin terns specificto different which they live (Levine and MacNichol 1979; Lythgoe 1979; Lythgoe and Partridge1991). Fish livingin clear water (tropicalmarineand shallow freshwater) tendto be moreblue sensitivethanfishlivingin colored (green,brown,or reddish ["black"]) waters, and fishlivingin deep water(but stillin the photiczone) tend to be less sensitive to reds than shallow-waterspecies (Levine and MacNichol 1979). It is likelythatthe conditionsunderwhichsensorysystemsare used affect the evolution of those systems. Sensory conditionsalso affectthe evolution of signals because they directlyaffectthe formof received signals (Endler 1990, themthroughtheevolutionof sensorysystems 1991)as well as indirectlyaffecting (fig. 1). The sensorysystem's characteristicsalso affecthow food is detectedand perceived. Food items may be most easily detected at particulartimes and places, which will affectthe evolution of microhabitatchoice, timing,and season of activityof predatorsand prey. These changes will in turnaffectthe evolutionof the sensorysystem(fig.1). The detectionof predatorsmay depend on microhabitat choice, and this may also affectthe evolutionof sensorysystems. Figure 1 summarizes the major evolutionaryinteractionsof sensory drive. There are a few intimatelyrelated processes that are also important,and these specializationis likelyto lead are shown in figure2. First, microenvironmental to changes in the visibilityor detectabilityof food (v). Food detectabilitycan processes: (1) directeffectson conspicuousness change as a resultof two different and the sensorysystem caused by the interactionbetweenthe microenvironment and (2) natural selection on the food organismto minimizeits visibility.These processes can result in food specialization and furtherhabitat specialization, whichin turncan affectsensorysystemevolution.Feeding success (fs) mediated by food visibilitycan also have directeffectson theevolutionof sensorysystems. specializationaffectthe visibilityor detectability Changes in microenvironment (p) of the signals to predators,which also affectsthe signal design. As forfood visibility,thereare two processes arisingfrompredation:(1) directeffectscaused by interactionsbetweenthe environmentalconditionsand the predator'ssensory systemand (2) naturalselection actingon the predatorto increase efficiencyin Both predatorsand food will affectthejoint evolutionof thatmicroenvironment. behavior, sensory systems,and signals. Mating success and sexual selection also affectthe process of sensory drive (fig.2). Mating success (ms) can affectthe evolutionof sensorysystemsdirectly ifmates are hard to detector courtshipritualsare verycomplex. Individualswith bettersenses may be more successful in findingmates than those with poorer senses, whichwould favorthe evolutionof sensorysystems.In addition,sensory power may be able to assess (and cognitive)systemswithmore signal-processing mates more rapidly,which would reduce risk of predationduringcourtshipor allow more matingsper lifetime. Sexual selection (ss) by female choice directlyaffectsthe evolution of matechoice criteria(fig.2), in additionto sensorydrive. The directionof the Fisher process of sexual selection is unbiased withrespectto the directionof evolution This content downloaded from 152.14.136.96 on Mon, 3 Jun 2013 09:12:05 AM All use subject to JSTOR Terms and Conditions SIGNALS AND THE DIRECTION OF EVOLUTION S129 Sensory system characteristics Environmental fs Mns conditions duringsignal reception v Visibility of food Mate choice criteria Signalingbehavior: Male trait choice, microhabitat properties timingand season of courtship, etc. Vsibility to predators FIG. 2.-Sensory drive and otherintimatelyrelatedprocesses. Thickarrowsare as in fig. on the 1; thinarrowsindicateancillaryprocesses. v, Immediateeffectof microenvironment visibilityof prey, also natural selection acting on prey; fs, feeding success that directly affectstheevolutionof sensorysystems;ms, matingsuccess thatmightaffecttheevolutionof sensorysystemsdirectly;ss, sexual selection(good genes or Fisher processes) thatdirectly on influencesthe evolutionof mate-choicecriteria;p, immediateeffectof microenvironment conditions visibilityto predators,and also naturalselection caused by microenvironmental actingon predatorsenses and behavior. and requires some externalfactorto set its initialdirection(Fisher 1930; Lande 1981; Kirkpatrick1982). Sensory drive can easily set the initialdirectionof the Fisher process and thus can profoundlyaffectthe directionof sexual selection. Various "adaptive," "good genes," or "handicap" systemsof sexual selection fitness(Pomiankowski favorkinds of traitsthatare good predictorsof offspring 1988; Grafen1990), but the exact traitsused may be determinedby otherfactors, such as sensorydrive. For example, carotenoid-basedyellow and red colors may both indicatefeedingsuccess, but visual conditionsand spectralsensitivitymay favorthe use of red. Sexual selectionby intermalecompetitionshould also operate simultaneouslywithsensorydrive; signalsthatmales use to assess each other are subject to the same biophysicaland neurobiologicalrules as other signals. for sexual selection to operate withoutsensorydrive, and It would be difficult vice versa. The major point of sensorydrive, and of figures1 and 2, is thatthe evolution This content downloaded from 152.14.136.96 on Mon, 3 Jun 2013 09:12:05 AM All use subject to JSTOR Terms and Conditions S 130 THE AMERICAN NATURALIST of sensory systems, signals, and behavior is coupled; changes in one suite of traitscause evolutionarychanges in the others. These suites of traitsmust not be expected to evolve independently;theywill coevolve. The evolutionof these traitsshould not be random, but it should, in principle,be predictablefroma knowledgeof environmentalphysicsand the biophysicsand neurobiologyof signal transmissionand reception.Particularenvironmental conditionsfavorparticular sensorycharacteristics.These characteristicsfavorspecificmate-choicecriteria, which in turnfavorcertainsexual signals. These signalsare best sent under unique environmentalconditions;thus, cycles of specializationof each suite of traitsare established. A similarargumentcan be made forthe traitsinvolved in otherkinds of signals. SIGNAL CONTENT, STRUCTURE, AND CONSTRAINTS Natural selection does not favoronly signalsthatare efficient in transmission and reception. Guilfordand Dawkins (1991) make an importantdistinctionbetween the strategicand tactical "design" of signals. A signal's tactical design refersto its structureand efficacy:naturalselectionfavorssignalsthatare easily transmitted,received, detected, and discriminatedfromothers,and those that are rememberedmost easily. For example, a particularsound patternmay be mosteasily heard underbackgroundnoise conditionsdeficientin thatfrequency, and it may be easy to distinguishfromothersignalsin the same frequencyband if its temporalpatternis complex and unique. In contrast,a signal's strategic designrefersto its purpose: naturalselectionfavorssignalsthatelicita response in thereceiverthatincreases or maintainsthefitnessofthesender.Thus, strategic design concerns signal contentmore than signal structure.Signals may contain true or false informationabout distance among neighbors(Morton 1982), social status,mate quality,and even the intentionsof the signaler(Sebeok 1977; Halliday and Slater 1983; Pomiankowski 1988; Grafen 1990; Guilfordand Dawkins 1991). Natural selection obviously affectsboth contentand structure,but these oftenresultfromdifferent processes. Sensory drive is primarilyconcernedwith the structureor tactical design of signals; it does not address the content or strategicdesign of signals, althoughit may provide constraintsas to what kinds of information can be sent. Most of the theoreticaland empiricalliteratureon sexual selection is more concernedwithsignalcontentthanstructure.This is particularlytrueof the good genes, adaptive, or handicap approaches, in which females use signal content as a predictorof offspringquality (Pomiankowski1988; Grafen 1990). In these approaches, the design constraintsinduced by sensorydrive may limitor bias the kind of informationthat can be sent; for example, less informationcan be sent per unit of time under noisy conditions.Design constraints,however, can also be used to convey information.For example, if all territoryholders send signals with the same sets of characteristics,and different parts of the signal degrade at differentrates with distance, then the configuration of the received signal, in comparison with the "expected" signal, contains informationabout distance (Morton 1982) and also, perhaps, rate of movement.Such information This content downloaded from 152.14.136.96 on Mon, 3 Jun 2013 09:12:05 AM All use subject to JSTOR Terms and Conditions SIGNALS AND THE DIRECTION OF EVOLUTION S131 could help the receiverdecide whetherto approach, especially iftravelcosts are high. If higher-qualitymale territoryholders are found at greaterdensity,the same informationmightalso be used to assess mate quality. The Fisher models of sexual selectionsimplyassume thatfemaleschoose mates on the basis of the male trait(the received signal). If the Fisher process works by itself,the directionof sexual selection is not inherentlybiased relative to content,and sensory drive sets the directionof evolution of the traits. Sexual selection under the Fisher process is most likely to affectthe structureof the signals ratherthan their content because it favors greaterefficiencyof signal receptionand processing(whichcan also be favoreddirectlyby naturalselection). Sexual selection under one of the good genes processes affectsboth structure and content.In the lattercase, sensorydrive affectsthe signalefficacyand constrainsand biases the directionof the evolutionof signalcontent. AlthoughI concentrateon signal structurein the rest of my article, I do not mean to imply that contentis unimportant;in fact, both strategicand tactical factorsshould always affectthe evolution of signals (fig.2). In addition,signal structurewill also evolve to be more efficientrelativeto events occurringin the brain. However, because littleis actuallyknownabout the psychologicalaspects of communicationand these aspects have alreadybeen exploredby Guilfordand Dawkins (1991), I consider only the processes occurringbetween transmission and reception. Genetic, phylogenetic,and physiologicalconstraintsalso affect the evolutionof signals,receptors,and behavior(i.e., in vision; Goldsmith1990); however, except forthe biases theymay induce in the directionof evolution,I do not discuss constraintsin this article. Again, this does not implythat these processes are unimportant, only thatthe purpose of thisarticleis to explore the effectsof signal design, reception,and associated behavior on the directionof evolutionin these suites of traits. A POSSIBLE EXAMPLE OF SENSORY DRIVE: VISUAL SIGNALS IN GUPPIES The System The color patternsof guppies (Poecilia reticulata)illustratea possible example of the results of sensorydrive for vision, visual signals, and behavior. Guppies are small poeciliidfishesof smallmountainstreamsfoundin thetropicalforestsof northeasternSouth America. They are geneticallypolymorphicforcolor-pattern elementsthat vary in color (hue), brightness(total reflectance),size, and shape. The color patternsin a given populationrepresenta compromisebetween sexual selection for conspicuousness and naturalselection for crypsis (inconspicuousness), and the compromisevaries geographicallywithpredationintensity.A male guppy cannot be too conspicuous, or it will be eaten beforeit has a chance to mate, but it cannot be too cryptic,or it will not be attractiveto females. Female preferencesvary geneticallyamong populations(Houde 1988b; Stoner and Breden 1988; Houde and Endler 1990), but, in general,femalesprefermales that are more conspicuous (Endler 1980, 1983; Kodric-Brown 1985; Houde 1987, 1988a, 1988b; Long and Houde 1989). Guppies vary withregardto the spectral This content downloaded from 152.14.136.96 on Mon, 3 Jun 2013 09:12:05 AM All use subject to JSTOR Terms and Conditions S132 THE AMERICAN NATURALIST sensitivityof theirlong-wavelength-sensitive cones (Archer et al. 1987), which suggests a mechanism for variationin female preferences.If this variation is heritable,then directnaturalselectionof the visual systemis possible. Guppies live in the company of a varietyof combinationsof diurnalvisuallyhuntingfish and invertebratepredators,and thisgeographicallyvaryingcommunitystructure resultsin predation-intensity gradientswithinstreams.As diurnalvisuallyhunting predationincreases in space or time,those fishwithgreateraverage crypsisare favored (Endler 1978, 1980, 1983). There are differencesin the timingof courtship relative to predation(Endler 1987), differencesin visual conditionsduring courtshipand predation,and differencesin vision among guppies and some of their predators (Endler 1991). These factors reduce the compromisebetween sexual selection and crypsis, because they cause the same color patternto be relativelyconspicuous undersome conditionsand relativelycrypticunderothers (Endler 1991). AmbientLight and Its Effectson Visual Signals In order to describe sensorydrive in guppies, it is firstnecessary to describe the environmentalconditionsduringsignalingand the effectof the ambientlight on the appearance of the signal. These conditionsand effectswill also be used forthe generalpredictionsabout visual signals. Guppies live in streams that flow throughtropical forests,which exhibit a mosaic of light environments(fig. 3). There are five differentlight environments-forest shade, woodland shade, small gaps, large gaps, and early/lateand these are determinedby forestgeometry,sun angle, and weather (J. A. Endler, unpublishedmanuscript).Forest shade results when most of the light reachingthe surfaceof an animal or backgroundhas been transmitted throughor reflectedfromvegetation;very littlecomes fromthe sky throughcanopy holes, and none comes directlyfromthe sun (fig.3). Forest shade is green or yellowreflectedand transmitgreenbecause intermediatewavelengthsare differentially ted by vegetation. Woodland shade results when a significantfractionof light comes fromthe open cloudless sky, the rest comes fromvegetation,and none comes directlyfromthe sun. Woodlands are forestswithlargelydiscontinuous canopy. Woodland shade is bluish or blue-gray,because the blue sky lightoverwhelmsthe lightreflectedfromthe vegetation.This lighthabitatis not limitedto forestsand occurs anywherethereis shade but where most of the sky is unobstructedby vegetationor clouds; thisincludes the shade of canopy emergentsin a forestcanopy and even the shade of boulders in the desert. Small gaps are patches of lightwitha diameterless thanabout 1 m (resultingfroma canopy gap subtendingless thanabout 1?solid angle). Small gaps tendto be relativelyyellowish or reddishcompared to fullsunlightbecause virtuallyall of the ambientlight comes directlyfromthe sun; the sun is muchricherin long wavelengthsthanthe sky or vegetation.Large gaps yield "white" lightand have essentiallythe same color as open areas because virtuallyall of the ambientlightcomes fromthe sun and open skyand littlefromthe vegetation.Clouds are whiterthandirectsunlight because theyare simplediffusersof the blue sky and sun above them.When the sun is blocked by clouds and a large fractionof the sky is cloudy, then forest This content downloaded from 152.14.136.96 on Mon, 3 Jun 2013 09:12:05 AM All use subject to JSTOR Terms and Conditions SIGNALS AND THE DIRECTION OF EVOLUTION Sun Sun Early/Late Cloudy 400 500 600 700 400 500 600 700 Forest Shade 400 500 600 700 S133 Small Gaps 400 500 600 700 Woodland Shade 400 500 600 700 Large Gaps 400 500 600 700 FIG. 3.-The major lightenvironmentsof forestanimals. The spectra shown are plots of the intensityof lightas a functionof wavelength(Endler 1990). Forest Shade is rich in middlewavelengths(green and yellow); Small Gaps are richin longer(redder)wavelengths; Woodland Shade is richin shorter(blue) wavelengths;Large Gaps exhibitessentiallywhite light.When the sun is obscured by clouds (Cloudy), the spectra of these fourhabitatsconverge on that of large gaps or nonforestedareas. In twilightconditions(Earlv/Late), the spectra of these habitatsconverge on a purplishlight,deficientin middlewavelengths.If a forestcanopy has taller emergenttrees, then both woodland shade and large-gapenvironmentswill be foundin the canopy. shade, woodland shade, small gaps, and large gaps convergein color and are all "6white,"as long as thereare some holes in thecanopy (fig.3, inset).This convergence occurs fortwo reasons: clouds are brighterthanblue sky and the radiance fromvegetation,so they overwhelmthese effects;and clouds are diffusers,so they increase the lightcoming throughall canopy holes in all directions.As a result,heterogeneityin ambientlightcolor disappears as soon as the sun is obstructedby a cloud. For thisreason I referto the white-coloredhabitatas "large gaps/open/cloudy,"or simply,"open/cloudy." A fifthlighthabitat,early/late, appears when the sun is risingor setting,regardlessof the weather.It is purplish (fig.3, inset) because at those times intermediatewavelengthsare differentially absorbed by ozone over the long light-pathlengths.If thereare clouds at those times,theremay be a briefperiodof stronglyreddishlightcaused by thereflection of long wavelengthsoffthe clouds. In summary,dependingon location, clouds, and time of day, a guppy (or any otherforestanimal) can be seen in greenish, bluish, reddish,whitish,or purplishlight(forestshade, woodland shade, small gaps, open/cloudy,and early/late,respectively;fig.3). The color (spectral shape) of lightreachingthe viewer's eye depends on the ambientlightstrikingthe color pattern,the reflectancespectrumof the colorpatternelement(colored patch), and the spectraltransmissionpropertiesof the wateror air (Endler 1990). For example, a color patternconsistingof gray,blue, This content downloaded from 152.14.136.96 on Mon, 3 Jun 2013 09:12:05 AM All use subject to JSTOR Terms and Conditions S 134 THE AMERICAN NATURALIST yellow-green,and red patches shows highcolor and brightnesscontrastin white light(open or cloudy conditions;fig.3), but underthe yellow-greenlightof forest shade the yellow-greenis very bright,whereas the blue and red patches are the greatest darkerand duller. In woodland shade the blue is brightest(reflecting proportionof ambientlight),whereas the othercolors are duller. In small gaps thered patches are brightest,and theblues are thedullest.In general,thepatches whose reflectancespectra match the ambientlightspectrumare the brightest, and those patches that mismatchare the dullest, for a given total reflectance (Endler 1986, 1990, 1991). This affectsboth the brightnessand color contrastof adjacent patches, thus affectingthe overall conspicuousness of the color pattern absorp(Endler 1991). The overall contrastis also affectedby wavelength-specific tion and scatteringof lightbetween the animal and the viewer. For example, in greenish water, blue and red patches attenuatemore rapidlythan yellow and betweenthe radiance specgreenpatches. Therefore,dependingon the similarity tra of each patch and the water (or air) transmissionspectra, the contrastcan change markedlywithdistance (Lythgoe 1979; Endler 1986, 1990, 1991). Hence, even ifthe animal's pigmentpatternremainsconstant,ifit moves amongdifferent distances, its appearance can lightenvironments,and if it is seen at different change dramatically(Lythgoe 1979; Endler 1986, 1990, 1991). By choosing when visual signals and whereto court,a guppyor otheranimalcan send verydifferent environments(Endler 1991). in different SensoryDrive in Guppies The followingsituationis hypotheticalbut consistentwithall our knowledgeof guppies. Consider an ancestralguppypopulationwithall color-patternelements equally frequent(i.e., yellow is as commonas blue and orange). A middaypeak in predationfavorsmost courtshipearly and late in the day and less courtshipat lightmidday(Endler 1987). Consequently,courtshiptakes place underdifferent ing conditions than those of maximumpredationtimes (figs. 2, 3). There can diurnalhabitats(fig. 3) affectcolor also be a spatial shiftbecause the different conspicuousness in various ways (Endler 1991). Courtshipand sexual selection favorcolor patches thatreflecta greaterfractionof the incidentlightduringthe times and places of maximumcourtshipactivityand select against colors that reflectless lightat those times and places. In general, the more lightreflected, the greater the potential visual contrast. Because maximumcourtshipoccurs under purplishlight (rich in blue and red light; fig. 3), blues and oranges are favored,while yellows are at a disadvantageat thistimeof day. Predationselects againstcolors thatreflectrelativelymore thanothercolors duringtimesof maximum predation risk, and this makes yellow particularlydisadvantageous. The conflictingeffectsof visibilityto both mates and predatorsfavor patches with maximumreflectanceand contrast duringcourtshiptimes and relativelylittle reflectanceand contrast duringpredation. These differencesare enhanced by differencesin vision and viewingdistances between guppies and theirpredators (Endler 1991). Visual differencesfavorcolors thatsignalto otherguppies at "private wavelengths," or at least those wavelengthsto whichpredatorssuch as the pike cichlid Crenicichla alta or the prawn Macrobrachiumcrenulatumare not This content downloaded from 152.14.136.96 on Mon, 3 Jun 2013 09:12:05 AM All use subject to JSTOR Terms and Conditions SIGNALS AND THE DIRECTION OF EVOLUTION S135 verysensitive(Endler 1978, 1991). As the color patternsbecome more and more efficient forcourtshipand less visible to predatorsat particulartimesand places, thereis selectionformicrohabitatchoice, because courtship,feeding,and other activitiesat other times and places result in lower matingsuccess and greater predationrisk.At timesofpredationrisk,guppiesbecome less visibleto potential mates and more visible to predators(Endler 1991). All of these factorsinduce at particular naturalselectionactingon the visual systemforincreasedefficiency timesand places. The neteffectis thatthe sensorysystemand sensoryconditions can affectthe visibilityof male traitsand, hence, mate-choicecriteria(through visibilityeffects),color-patterndesign, and microhabitatchoice and timingof courtship;these factors,in turn,can affectthe evolutionof the sensorysystems. This leads to a cycle of evolutionaryinteractions,as shown in figure2. SOME GENERAL PREDICTIONS FOR VISUAL SIGNALS The example of sensory drive in guppies and the general scheme shown in to demonstratedirectly.Yet it is possible to predict figure2 would-be difficult how suites of sensory, signal, and behavioral traitsshould jointly evolve and, hence, be distributedin nature. The predictionsin the next section apply to conspecificvisual signals and aposematic signals but not to crypsis. Note that these predictionsare tentativeand qualitative; specificand quantitativepredictions would require a detailed knowledge of the vision, signals, environmental physics,and behavior of the animals concerned. MaximizingConspicuousness In a particularlightenvironment,a color patternis most conspicuous if its adjacent color-patternelementsvary greatlyin brightness(total reflectance)and chroma (saturationor color purity),and it is less conspicuous if it varies less in these parameters(Endler 1990, 1991). Spectra with high saturationhave rapid transitionsin intensityas a functionof wavelength,whereas those withlow saturation have only gradual transitions(Endler 1990). Unsaturated (low chroma) colors change withambientlightmore than saturatedcolors; a perfectlyunsaturatedspectrum,such as whiteor silver,simplyreflectsthe same colors thatstrike it. There are fourways to maximizecolor-patternconspicuousness. The simplest methodis to have color patternswith lightpatches (white, lightgray, or other highlyreflectiveunsaturatedcolors) adjacent to dark patches (black, dark gray, or others with low reflectanceand chroma). But color patternswith adjacent unsaturatedcolors of contrastingreflectanceare disadvantageousin thattheyare conspicuous in nearlyall lightconditionsto nearlyall species, so any advantage in courtshipmay be offsetby increased predation. The second methodof maximizingconspicuousness is to matchthe brightness of the ambientlightand water (or air) transmissionspectra. If the ambientlight wavelengths strikingan animalis notwhite,thenitconsistsof some high-intensity and otherlow-intensity wavelengths.To be conspicuous, a color patternshould ambientwavelengthsand these have colored patches that reflecthigh-intensity This content downloaded from 152.14.136.96 on Mon, 3 Jun 2013 09:12:05 AM All use subject to JSTOR Terms and Conditions S136 THE AMERICAN NATURALIST patches should be adjacent to patches thatreflectlow-intensity wavelengths.For example, in bluish light,a blue patch would reflectmost lightstrikingit (bright spot), whereas a yellow or red patch would reflectvery littlelight(dark spot), and a green patch would be intermediate,assumingthe same total reflectance. The thirdmethodof maximizingconspicuousness is to have adjacent patches withcomplementarycolors, such as red and greenor yellow and blue. Complementarycolors are colors whose spectrahave fewor no wavelengthsin common, for example, blue and yellow. The radiance spectrumof a blue spot is rich in wavelengthsbelow 500 nm but radiates littlelightabove 500 nm, and a yellow spot radiates much above 500 nm but littlebelow 500 nm. Adjacent patches are the most conspicuous because theircomplementaryradiance spectra stimulate combinationsof wavelength-specific photoreceptors(cones) in opposite ways, whichmaximizescolor contrast(Lythgoe 1979;Hurvich1981).This maximization of contrastdepends on the peak absorbance wavelengthsof the cones relativeto the cutoffwavelengths(the wavelengthat which radiance changes fromlow to high,forexample, at about 600 nm foran orangepatch), as well as on the details of the neural connectionsin the retinaand the brain. Thus, two patches thatare complementaryfor one species may not be complementaryforanother,and so theywill seem less brightto the latterspecies as comparedto the former. The fourthmethod of maximizingconspicuousness is to use complementary colors whose cutofffrequenciesare centeredin theregionof thegreatestambient lightintensityand water (or air) transmission.For example, in the bluishlightof tropicalmarinewater,blue and yelloware moreconspicuous thanred and green, whereas, in the greenishlightof temperatelakes and ponds, red and green are more conspicuous thanblue and yellow. This factmay explain the relativeabundance of these pairs of colors in tropicalmarineand temperatelake fishes,respectively (Lythgoe 1979). As in the case of brightnessmatching,any change in ambientlightand/ortransmissionspectrawillchangetherelativecontrastofpairs of colors thatare complementaryat different parts of the spectrum. Divergence of Species and Populations Closely related species thatlive in different lightenvironmentsshould exhibit the followingdifferences:predictablydifferent color-pattern characteristics,predictably differentsensory characteristics,and predictablydifferentbehavior. These differencesmay also apply among populations of widespread species if they vary geographicallyin microhabitatand lightenvironment.Because there are fivemajorlightenvironmentsin forests(fig.3), thesepredictionscan be made more specific.(These predictionsare based on the fourmethodsof maximization of contrastand conspicuousness describedin the previous section.) Color-patterncharacteristics.-Species that live in forests with continuous canopy and thatcourtin forestshade (fig.3) should use primarilyred and orange and a littleblue forincreased contrastwithinthe color pattern.Species thatlive in woodlands or in thecanopies offorests(geometricallyequivalentto woodlands) and thatcourt in shade (woodland shade; fig.3) should use primarilyblue, bluegreen,and, ifpossible, ultravioletpatches, witha littlered forincreasedcontrast. Species that court in small gaps should use yellow or orange, with some purple This content downloaded from 152.14.136.96 on Mon, 3 Jun 2013 09:12:05 AM All use subject to JSTOR Terms and Conditions SIGNALS AND THE DIRECTION OF EVOLUTION S137 forincreased contrast.There are no particularpredictionsforspecies thatlive in large gaps or court anywherein forestsor woodlands when the sun is obscured by a cloud (open/cloudy; fig. 3), because all wavelengthsare roughlyequally abundant;therefore,no colors would be moreefficient thanothersin maximizing the signalconspicuousness. If one were to compare the color patternsof species livingin evergreentropicalforeststo those of species livingin deciduous tropical forests,one would also expect to find,on the average, more use of ultraviolet, blue, and green in seasonal forests,because woodland shade would be more commontherethanforestshade. Duringthe dryseason thereis verylittlecontinuous canopy (fig. 3) in tropical deciduous forests,but the canopies of tropical evergreenforestsremain relativelycontinuousall year. Similarly,species that breed in the dry season should use blue and green in theirvisual signals more oftenthan those that breed in the wet season. One has to know exactly where the species display when one teststhese predictions.For example, the lightenvironmentat or near forestcanopies is essentiallythe same as woodland: the canopy emergentsforma discontinuouscanopy resultingin a mosaic of woodland shade and large gaps immediatelybelow the emergents.Consequently,we would expect a greaterfrequencyof blue and green in canopy species than in forest floorand subcanopy species, all in forestswithcontinuouscanopy. Because all forestlightenvironmentsconverge on large gaps or open areas duringcloudy weather(fig.3), these predictionsshould have greateraccuracy in places and times withfewercloudy days and should be least accurate in cloud forests.Because the relativelyflatambientlightspectraof cloudy days in forests do not favor any particularcombinationof signalingcolors, there should be a greaterdiversityof signalingcolors in cloud foreststhan in other places with lower numbersof cloudy days per year. There are two commonpatternsin tropical rainforestsduringthe wet season. In some places (such as Costa Rica) heavy rains come in the afternoon,but it is frequentlysunnyin the morning.In other places (such as Trinidad) thereis no predictablepattern,and it may be cloudy all day. Thus, one would predicta greaterdiversityin the signalingcolors of wet-seasonbreedersin places wherethe cloudinessis unpredictableand common than in places where there are regularsunnyperiods in the wet season. These predictionsshould also be more accurate for species that are active only when and inactiveas the sun is out (tanagers,orioles, and many species of butterflies) soon as the sun is obstructedby clouds. But theywill be invalidforspecies that are usually active when the sun is obscured by clouds. Visual characteristics.-The characteristicsof the sensory system should matchthose of theenvironmentin orderto make themostuse of available signals. Species displayingin forestshade should be relativelymore sensitiveto yellowgreenand perhapsred thanspecies displayingin theopen or in largegaps. Species displayingin small gaps should be relativelymore sensitiveto longerwavelengths (yellow, orange, and red). Species displayingin woodland shade should be relativelymore sensitiveto blue and blue-green.In general,species foundin continuous forest(forestshade and small gaps) should be more sensitiveto longerwavelengths,whereas species found in woodlands (woodland shade and large gaps) should be more sensitiveto shorterwavelengths.Once again, canopy species in This content downloaded from 152.14.136.96 on Mon, 3 Jun 2013 09:12:05 AM All use subject to JSTOR Terms and Conditions S 138 THE AMERICAN NATURALIST forestsshould be more similarto woodland species than eitheris to forestfloor and subcanopy forestspecies. These predictions,like those forsignalingcolors, need to be modifiedto account forthe frequencyof cloudy weather. Behavior.-Behavior involvingvisual signalsto conspecificsor, ifaposematic, signals to potential predators should also vary with the environment.Signals should be sent at the times and in the places and seasons in which theirsignalnoise ratiowill be maximized.For example,ifa birduses blue to signalto conspecifics,it should do so in the shade of canopy emergents(effectivelywoodland shade) ratherthan in the dense shade founddeeper in the forest(forestshade), and it should not be as active when the sun is obstructedby clouds. The need to the signalsto predatorsalso affectsplaces signalto conspecificswhile minimizing in whichthe animalforages.Thus, species foundin shadymicrohabitatsin forest canopies (woodland shade) are also found in woodland habitats,the edges of tree falls, and disturbedsites, whereas species foundin shady habitatsbeneath continuouscanopies (forestshade) are much more habitatspecific.By the same logic, species specializing in small gaps may have a large heightrange within forestsbut probablywill not go outside dense forests.The same argumentsmay apply also to singleisolated shrubsforsmallerinsects,because the lightenvironmentvaries withgeometryin the same way as in forests,but on a much smaller scale. Species thatsignalwhen in the sun should have a greaterrangeof habitats and microhabitatsthan those signalingin one of the shade microenvironments. Similarly,species that are active duringcloudy conditionsshould also have a greaterrange of habitatsand microhabitatsthanthose thatare active only when the sun is not blocked by clouds. The evolution of behavior and color patternsshould interact.If behavior is indeed selected to maximize the efficiencyof signal transmissionand reception to conspecificswhile minimizingthe signals to potentialpredators(fig.2), then thereare certaincharacteristiccolor-patterndesigns that should evolve repeatsignals patternto send different edly in orderto allow the same body-reflectance at different times and underdifferent conditions. One way to vary signals using the same color patternis to take advantage of the relationbetween brightnessmatchingand contrast.Brightnessmatchingof theambientlightand transmissionspectracan increasecontrast,butthisdepends on the spectra of the animal's color-patternpatches. For example, as mentioned earlier,in bluishlight,a blue patch would be brightwhereasa yellowor red patch ifone assumes the same would be dark,and a greenpatchwould be intermediate, The relative radiance of the three totalreflectance. patches would shiftmarkedly undergreen, yellow, or red lightand, consequently,so too would the color pattern's brightnessand color contrast (Endler 1986, 1991). The effectcan be strongerif the color patternconsists of both saturatedand unsaturatedcolors. The radiance of unsaturatedcolors changes withambientlightmore rapidlythan thatof saturatedcolors (fig.4). This allows thedegreeof conspicuousnessto vary with lightconditions. The transmissionpropertiesof colored water or foggyor dusty air may also be taken advantage of in order to vary the conspicuousness of a singlecolor pattern.Displayingat shortdistancesto conspecificswhilebeing seen at longerdistances by predatorsand usingcombinationsof colors such that This content downloaded from 152.14.136.96 on Mon, 3 Jun 2013 09:12:05 AM All use subject to JSTOR Terms and Conditions SIGNALS AND THE DIRECTION OF EVOLUTION S139 1.0 IL 0.8Gray 0.6----0Iui LL. - 0.4- Wu C - - -- - -- - Orange 0.20.0 400 450 500 550 600 650 700 WAVELENGTH(nm) FIG. 4.-An example of how the color of ambientlightaffectsthe contrastbetween two different patches, in this case, a brownishgray and an orange patch. If both patches were illuminatedby white light (same intensityat all wavelengths),the gray patch would be brighterbecause it reflects1.71 times as much lightas the orange patch; the orange patch reflectsvery littlelightbelow 550 nm. If both patches were illuminatedby orange light(no lightat 400-550 nm,flatspectrumat 550-700 nm), the relativebrightnessof the two patches would be very similar(0.98: 1.00), and an orange patch would disappear on the gray background (everythingwould appear orange to an animal with color vision). This is simply because the reflectancesof the two patches are similarin the available light.However, if both patches were illuminatedin greenishlight(flat spectrumat 400-550 nm; no lightat 550-700 nm), therewould be veryhighbrightnessand color contrastbecause the graypatch would reflect8.74 timesas muchof theavailable lightas would theorangepatch. The orange the available lightif the available lightis orange or red. patch is most efficientin reflecting Similareffectsare foundforothercolors; patches are most efficient as signalsiftheymatch the ambientlightspectrum. the patternhas lower contrastat longerdistancesalso allow behavioralmanipulation of the appearance of color patterns,as in guppies (Endler 1991). Anotherway in whichconspicuousnessmayvaryis ifthecolor patternconsists of colors easily seen by mates but not seen or relativelypoorlyseen by predators. For example, blue patches of guppies can be seen by other guppies but are probablynot as conspicuous to Crenicichla,and orange patches can be seen by guppies but are probably not as conspicuous to prawns (Endler 1978, 1991). Because there is so much variation in spectral sensitivityamong fish species (Levine and MacNichol 1979; Lythgoe 1979) and among other vertebratesand invertebrates(Jacobs 1981; Laughlin 1981; Goldsmith1990), theremay be ample opportunityfor private wavelengthsthroughoutthe animal kingdom.As mentionedearlier,the conspicuousness of complementary colors may also vary with the predator'svision. As in the case of conspicuousness, these methods are probably used by a wide varietyof species, but comparativestudies have not yet been made. The predictionsabout visual signalsreveal onlythe "tip of the iceberg" regardingthe This content downloaded from 152.14.136.96 on Mon, 3 Jun 2013 09:12:05 AM All use subject to JSTOR Terms and Conditions S140 THE AMERICAN NATURALIST TABLE 1 RULES FOR SIGNALING WITH SOUND 1. Use lower frequencies(<2 kHz) in orderto minimizereverberation,attenuation,and scattering by the atmosphere,objects, and "shadows" resultingfromtemperaturegradients.Do not use and to avoid enerfrequenciesthatare too low (>0.5-1 kHz) to minimizedestructiveinterference geticallycostly couplingbetween low-frequencysound generationand the atmosphere. 2. Send signals froma positionthatis greaterthan 1 m above the groundand preferablyupwind fromgroundreflectionand of the receiver,in order to minimizedestructiveinterference effects. temperature-gradient 3. Avoid rapid syllable repetitionrates in places (such as forests)withstrongreverberation. 4. Use frequencymodulationratherthan amplitudemodulationbecause reverberationand turbulence modifysound amplitudemore thanfrequency. *5. Use redundantsignal structurein orderto average out backgroundnoise. *6. Use greateramplitudeand/orphysicalstructuresto beam the signal more effectively. in a shorttimebecause remustbe transmitted *7. Use higherfrequenciesif much information ceptors have a fasterfrequencyresponse at higherfrequencies,and therewill be less degradation by turbulence. *8. Use species-specificfrequencybands and tunedreceptorsin orderto minimizenoise at other frequencies. *9. Choose frequencybands, places, seasons, and timesof day thatminimizeturbulenceand/or backgroundnoise. *10. Avoid signalingat the same timeas immediateneighbors(unless jammingtheirsignals is desired). *11. Use rapidlydegradingsignalsforshort-distancecommunicationand slowly degradingsignals forlong-distancecommunication. *12. Use simpler,more effectivealertingsignalsto attractthe receiver's attentionbeforesending the main signal. NOTE.-Rules markedwithan asteriskalso apply to othersensorysystems.The table is modified fromWiley and Richards 1982; Capranica and Moffat1983; Brenowitz 1986; Okanoya and Dooling 1988; and Ryan 1988b. effectsof sensory drive on systemsrelated to visual signalingand would repay muchfurtherresearch. Additionalpredictionsabout color patternsare discussed in Endler (1978, 1984, 1986). SOME GENERAL PREDICTIONS FOR ALL SENSORY SYSTEMS Introduction:SensoryDrive for Sound Sensory drive (fig.2) should apply to all sensorysystemsand signals,not only to vision and color patterns.For example, thereis enough known about sound and hearingto make some generalrules about the optimumdesign and timingof (terrestrial)auditory signals; these are summarizedin table 1. There is good evidence forthe operationof these rules in crickets(Jones 1966; Greenfield1988; Simmons 1988), frogs(Capranica and Moffat1983; Ryan 1988a, 1990; Brush and Narins 1989; Ryan and Rand 1990; Narins 1992; Ryan and Keddy-Hector 1992), birds(Henwood and Fabrick 1979; Richardsand Wiley 1980;Dooling 1982; Wiley and Richards 1982; Brenowitz 1986; Okanoya and Dooling 1988),mole rats(Heth et al. 1986), and primates (Waser and Waser 1977; Waser and Brown 1984). For detailed discussions, see Wiley and Richards (1982), Brenowitz(1986), and Okanoya and Dooling (1988). From this research,it is easy to envision sensory drive for sound: the auditorysystemfavors particulartraitsin male songs and This content downloaded from 152.14.136.96 on Mon, 3 Jun 2013 09:12:05 AM All use subject to JSTOR Terms and Conditions SIGNALS AND THE DIRECTION OF EVOLUTION S141 calls that are best used under particularconditions,times of day, and seasons. These traitsthen favor specialization of the auditorysystemfor use in those microhabitats,which in turn affectsmate choice and male sounds. A similar process should work in chemosensoryand electrosensorysystemsas well, and, indeed, many of the rules in table 1 (marked with an asterisk) apply to other sensorymodes. If sensorydriveworksin all sensorymodes, thenit shouldbe possible to make some general predictionsabout the evolution of signals, sensory systems,and behaviorby using the clues gained by a studyof visual and auditorysignals. The followingare some of the ways in whichsensorydriveshould affectthe direction of evolutionof these suites of traits.Once again, these are tentativeand purely qualitative predictions; detailed and quantitativepredictionsrequire detailed knowledge of sensory biology, environmentalbiophysics,signal structure,and behavior. Increasing Signal Intensity Signals should evolve to maximizethe signal-noiseratioforthe receiverunder the conditionsin which theyare sent and received. One of the simplestways of accomplishingthisis to increase the intensityor amplitudeof the signal(Capranica and Moffat1983). There is much evidence forthis in a varietyof organisms and sensorymodes; femalesusuallyprefermales withbrighter,more contrasting colors, fastermotion or flashingduringvisual display, louder sounds, stronger chemical or electricalsignals,and so on (see especially Ryan and Keddy-Hector 1992). Larger body size and tail lengthare also frequentlyfavored. Although otherexplanationsforthese preferencesare oftengiven,largertailsor bodies give a strongervisual signal,especially ifthe displayis accompanied by movement.It is probably safe to predictthat signals should evolve to become stronger,with constraintsset by the geneticsand energeticsof signalproduction(fora discussion of constraintsin sound, see Ryan 1988a, 1988b; fora generaldiscussion of limitsin any sensory mode, see Cohen 1984). However, if the signal-intensity signal is given against intensebackgroundnoise, naturalselection may favor a weak signal,and the "hole" in the backgroundnoise, which contraststhe signal and the noise, is the real signal. Black or dark animals against white or bright backgroundsare a good example, and thiskindof signalmay be used by crows, ravens, and vultures. The differencebetween the patternof light sent by an animal and the resultingvisual contrastbetween the animal and the background illustratesthatone needs to be carefulabout what is meantby "signal"; signals are context-dependent(Endler 1978, 1984, 1986, 1990). The signal-noiseratio can also increase throughthe evolutionof signaldesign. This is affectedby the physics of the environment throughwhichthe signalmust travel,the biophysicsof signalreceptionand processing,and the problemof not signalingto predatorsat the same time. Reducing Signal Degradation Natural selection to reduce signaldegradationand the physics of the environment favors particularkinds of signals with particularproperties.At the most This content downloaded from 152.14.136.96 on Mon, 3 Jun 2013 09:12:05 AM All use subject to JSTOR Terms and Conditions S142 THE AMERICAN NATURALIST basic level, the sensorymode thatevolves to predominatein signalingwill be the one withthe greatestefficiencyof transmissionin the local environment(Wiley and Richards 1982; Capranica and Moffat1983). For example, signalingis most efficientforvision by day in places where an unobstructedview is possible, for hearing in dense vegetation or at night,for electroreceptionin streams with cloudy water or nocturnalspecies, and forchemical communicationin burrows. Because signalsin all sensorymodes are degradedby the environment and during detection,if all else is equal, the sensory mode with the greatestpotentialfor information-transmission rate should be used in preferenceto modes with narrowerbandwidth.For example, visual signalscontaintemporaland spatial components,and the spatial component(at any one instantin time)varies in intensity and spectral composition. Thus, the added spatial componentmay mean that visual signals are able to transmitmore information than sound or electricalsignals, which contain only temporallyvaryinginformation.Chemical and some tactile signals may be able to transmitmuch information,but they cannot be frequencyor amplitude modulated as rapidlyas visual, auditory,or electrical signals. The choice of sensorymode or modes used in signals depends on both environmentaldegradationand potentialinformation-transfer rate. Using many sensorymodes is always betterthanusinga fewor one, but environmental conditions,energetics,and developmentalconstraintsmay limithow much each mode is actually used. Because these conditionsmay vary fromplace to place and among species, the use of multiple-modesignalingshould also vary. The relative importanceof differentsensorymodes can vary among species withina genus; for example, different combinationsof visual, Drosophila species use different auditory,and chemical cues duringcourtship(Ewing 1983). It should be possible to predictwhich mode (or modes) is most usefuland, therefore,which mode is mostcommonlyused, ifone knows thephysicsof theenvironment and thedesign of the signals. Withina sensorymode it should be possible to predictmanyof the details of the signals as well as the microhabitat,time,and season of transmission,if we know the environmentalphysics. The design should be one that attenuatesand degrades as slowly as possible, and it should be sent in the microhabitatand at the timeand season in whichdegradationis minimized.Many details were given earlier in this article for vision and hearing,and similarlogic could be used in other sensory modes. For example, in turbulentair or streams in which water of differentsalinities or lighttransmissionspectra are mixingturbulently, frequency-modulatedsignals are better than amplitude-modulatedsignals (for electric and sound signals, see Brenowitz 1986), and the signal sender should choose places and times withminimumturbulence.Electric signals are affected by watersalinity,so breedingelectricfishshouldavoid thewet season or particularly rainy days, when salinityis reduced (Brenowitz 1986). For any sensory system,signalingbehavior and microhabitatchoice can be used to maximizethe signal-noiseratio. If thereis geographicalvariationin degradationwithinthe range of a species, then its signals should vary accordingly.There are several examples. In sticklebacks (Gasterosteusaculeatus) thered nuptialcolor is muchrarerin places where This content downloaded from 152.14.136.96 on Mon, 3 Jun 2013 09:12:05 AM All use subject to JSTOR Terms and Conditions SIGNALS AND THE DIRECTION OF EVOLUTION S143 the water stronglyabsorbs red (Reimchen 1989). The geographicalvariationfor orange preferencein guppies (Houde and Endler 1990) may also relate to water color. A frogspecies and several bird species show the predictedsong variation withhabitatamongpopulationswithinspecies (Bowman 1979; Hunterand Krebs 1979; Gish and Morton 1981; Andersonand Conner 1985; Ryan et al. 1990), and an apparent exception may be explained by air turbulence(Wiley and Richards 1982). Two different populationsofDrosophila mojavensisshow different combinations of epicuticulardienes, which are used as pheromonesand affectmate choice (Markow and Toolson 1990). This is a particularlyinterestingexample because the volatilityof these hydrocarbonsis temperature-dependent and the bothfactorsaffectsignalefficiency. mean temperatureof thetwo sitesis different; In addition,thedienes affectwaterbalance, so thedirecteffectof signalefficiency and natural selection for water regulationmay jointly bias the directionof the evolutionof pheromonesin different populations(Markow and Toolson 1990). It would be interestingto know how common geographicalvariationin signals is withinspecies. There should also be a correlationbetween signals and habitat among species; this is reasonably well known for color patternsin fish(Levine and MacNichol 1979; Lythgoe 1979) and for sound, as mentionedearlier,and it would be interestingto know whetherit is trueforothersensorymodes. Reduction of Noise The signal-noiseratiocan be increasedby reducingbackgroundnoise. Behavior should evolve so thatsignalsare sent and received at places, times,and seasons withlow ambientnoise, which would otherwisemask or degrade the signal. As mentionedearlier,male guppies signalto femalesat a timewhen theywill reflect the most light and show the greatest visual contrastwithinthe color pattern (Endler 1991). An Anolis lizard must move its head and dewlap at a frequency fromthatof the movingvegetationbackground(Fleishman 1988, 1992). different The same considerationsapply to predators:the Anolis-eatingvine snake (Oxybelis aeneus) vibratesits body at approximatelythesame frequencyas theoscillationsof the vegetationbackgroundin orderto minimizedetectionby Anolis and, perhaps,the snake's own predators(Fleishman 1986, 1992). Noise can come from other species, which is a problem particularlywhen they are closely related (Brush and Narins 1989; Narins 1992). Because the conspecificsignalingtime is also adjusted to a timeof minimumenvironmental noise, the presence of conspecifics may cause a trade-offbetween reducingrandom environmentalnoise at certaintimesand reducingconspecificnoise at othertimes(Capranica and Moffat 1983). Neoconocephalus katydidsare a good example. In the absence of congeners, four species of katydidssing at night.However, Neoconocephalus spiza singsby day whereverit is sympatricwithone or moreof the otherthreespecies (Greenfield1988). At verylow signal intensitiesthe choice of when and where to signalbecomes even more important.For example, vertebraterods can react to singleabsorbed photons (Lythgoe 1979) but will firespontaneouslyeven in the dark, because of molecularnoise. At very low lightlevels, when rods are respondingto the occasional single photons strikingthem,random noise can significantly degrade the This content downloaded from 152.14.136.96 on Mon, 3 Jun 2013 09:12:05 AM All use subject to JSTOR Terms and Conditions S144 THE AMERICAN NATURALIST image (see photographsin Lythgoe 1979)because the spontaneousfiring rate may be similarto the signal (signal-noiseratio near 1.0). As the temperatureof the retinaincreases, the molecularnoise increases,whichmakes detectionof weaker signalsincreasinglydifficult. As a result,toads and frogsare actuallymore sensitive when theyare cold than when theyare warm(Aho et al. 1988). Because the chemistryof visual pigmentsis verysimilaramong vertebratesand invertebrates (Goldsmith1990), visual signals involvingthe lowest lightintensitiesmightalso occur at the lowest temperatures,such as at night(fireflies) and in the deep sea. The temperature-dependent signal-noiseratio applies to all sensorysystems,so we mightalso expect the weakest signals in any mode to be most common at nightand cooler seasons. Of course there are otherpossible reasons for these patterns.For example, signalingcan occur at nightas a resultof strongerdiurnal predation.Weaker signalscould occur at nightbecause thereis less environmental (ratherthan thermal)noise at that time; therefore,weaker and energetically cheaper signals are requiredto yield a given minimumsignal-noiseratio. The propertiesof visual signalsgeneratedby the animal(bioluminescence)can fromthose redirectedfromthe environment(color patterns). be quite different Most nocturnalfirefliessend a greensignalsimilarin spectralcompositionto the vegetationbackground(Seligeret al. 1982). This spectralmatchingcauses a larger fractionof the emittedlightto be reflectedoffthe vegetationthan if the flashes were whitewiththe same photonflux.If the flasheswere white,thenlongerand shorterwavelengthswould be differentially absorbed by the vegetation,which would yield a weaker signal. Firefliesthatflashat dusk have a different problem, because there is significantambient lightthat can interferewith their signals. However, at dusk the lightis purplish(deficientin middlewavelengths;fig.3). Firefliesthatflashat dusk minimizethe interference of ambientlightby flashing with a yellowishratherthan a greenishlight(Seliger et al. 1982). Their yellow flashesare rich in intermediatewavelengths,preciselythe wavelengthsthat are least intenseat dusk, makingthe signal-noiseratiomuchhigherthanit would be if they flashedgreen (Seliger et al. 1982; Endler 1991). Note the differencebetween the yellow flashes of dusk-signalingfirefliesand the rarityof yellow in the signalfroma male guppy guppies signalingat the same time. Unlike fireflies, originatesin the ambient light, so its color patternsmust match the ambient lightin orderto reflecta highproportionof it, whichthuskeeps a relativelyhigh signal-noiseratio. Signal Receiver Design A larger signal-noiseratio can also be achieved by specialized design of the signal receptorsand how the signals are processed afterreception.A common way to reduce noise is to adjust the characteristicsof the signal receptorsto matchthe characteristicsof the signal(Capranica and Moffat1983). This is most efficientif the noise exists over a broad band of frequencies(light,sound, or electricsignals) or chemicals. If a signalconsistsprimarilyof onlya limitedrange of frequencies,thena tuned receptorattainsa highersignal-noiseratio than one thathas a broad sensitivity.For example, fireflies thatflashyellow lightat dusk are particularlysensitive to yellow lightand not very sensitiveto other wave- This content downloaded from 152.14.136.96 on Mon, 3 Jun 2013 09:12:05 AM All use subject to JSTOR Terms and Conditions SIGNALS AND THE DIRECTION OF EVOLUTION S145 were sensitiveto greenlight,as are lengths(Seliger et al. 1982). If these fireflies theirnight-flashing congeners, then they would receive much irrelevantbackgroundlight(noise) and proportionallyless signalfromtheirflashes.This signal matchingis quite common in virtuallyall sensory modes in which it has been examined,which yieldsa good correlationamongspecies betweensignalcharacteristicsand receptorproperties.For example, receptorsignalmatchingis found in color patternsand vision in fishes(Levine and MacNichol 1979; Lythgoe 1979) and Lycaena butterflies(Bernard and Remington1991), in motiondetection in lizards (Fleishman 1986, 1988, 1992) and insects (Alexander 1962; Seliger et al. 1982), and in hearingin birds (Wiley and Richards 1982; Okanoya and Dooling 1988) and frogs(Ryan 1990). A correlationbetween signals and signal receptors shouldexist withinthe limitsof energeticand phylogeneticconstraints(Ryan and Brenowitz 1985; Ryan 1988a, 1988b, 1990), and the correlationshould be tighter forthose environmentsand those senses thatare associated withgreaterenvironmentalnoise. The signal tuningmay be set to minimizethe noise in the environment,butitautomaticallyfavorsthoseindividualswho send signalsbest matching is Ryan's (Ryan 1990; Ryan and Rand 1990) "sensory exploitathereceptors:-this tion" model (fig. 1, shaded areas). The evolutionaryresult is that signals are tunedto matchthe characteristicsof the receptors.Of course, the receptorscan also evolve (fig.2). Noise reductionby tuningneed not be restrictedto frequencybands but may involve nearly any aspect of signal structure;the only criterionis reductionof backgroundnoise. This is mostapparentin visual signals,whichare verymultidimensional.A varietyof vertebrateand invertebratevisual systemshave various respondto particukindsof edge detectorsand neural "units" thatpreferentially lar shape and size elementsof images and to motionin particulardirectionsand speeds (Ewert et al. 1983; Guthrie1983; Stone 1983; Blakemore 1990). Thus, any signals other than those that fitthe receptorcharacteristicsin great detail are automaticallyfilteredout. For example, fiddlercrabs and other semiterrestrial decapod crustaceansare verysensitiveto verticallyorientedobjects, and a vertical towerat the burrowof Uca beebei is a significant componentof femalechoice (Christy1988). Of course the pillars also allow fiddlercrabs to findtheirbeach burrowsin a hurrywhentheyare chased by predators.This "templatematching" between specific signal structureand neural units in the brain has also been suggestedforsound but so farhas only been demonstratedin foragingbats (Capropertiesof the pranicaand Moffat1983). It is possible thatthese signal-specific sensorysystemcan predispose the evolutionof femalechoice and male traitsin particulardirections(Basolo 1990; Ryan 1990; Ryan and Rand 1990; Ryan and Keddy-Hector 1992). visual propertiesmayalso have profoundimplicationsforhabiPattern-specific tat selection. For example, Morton(1990) foundthatmale hooded warblers(Wilorientthemselvestowardverticalstripesand prefer sonia citrina)preferentially artificialhabitatswithtall stemsratherthanthose withlow, dense stems.Females toward diagonal stripesand have no significant orientthemselvespreferentially artificialhabitatstructures.These differencesare associpreferencesfordifferent ated in the fieldwithmales' spendingmoretimein theforestand females' spend- This content downloaded from 152.14.136.96 on Mon, 3 Jun 2013 09:12:05 AM All use subject to JSTOR Terms and Conditions S 146 THE AMERICAN NATURALIST grounds(Morton ing more time in shrub or open habitatsin the overwintering 1990). These differencesare also associated with differencesin the mean and variance of male and female color patterns,which one would expect fromthe radically differentlightenvironmentsof the two habitats (fig. 3). For poikilotherms,particularlysmall ones, such as insects, strongmicrohabitatpreferences are associated witha host of biological consequences thattend to be associated microenvironments (Willmer1982), which furtherencourages difwithdifferent ferentiation.These are the kindsof differencesthatcould easily cause and maintain species differencesand could suggestways in which sensory systems,signals, and behavior can coevolve (fig.2). Signal Processing Special mechanismsof signal processingcan also increase the efficiencyof a signal. One way to circumventproblemsof both backgroundand sensor noise, especially at low signal intensities,is by averaginga signal over time. If noise fromthe environmentor the receptorsis random,thenaveraginga signalcauses the randomnoise fluctuationsin the received signalto cancel out; thus,a greater frequencyof signals with redundantand repetitivestructureshould evolve in sensorymodes and environmentswithgreaternoise. Certainlymany signals are repetitive,but thereare not yet enough data on environmentalnoise to test this prediction. Anotherfactoris sensory adaptation or habituation-the phenomenon,common to all senses, of ignoringconstantsignals. This may allow unpredictableand complex, yet monotonous,backgroundnoise to be filteredout. The commonness of sensory adaptation suggests that temporallyvaryingsignals should be more common than constantones. This is true even in complex visual signals; visual displays are usually associated withmotion.But sensoryadaptationmay require the use of alertingsignals. Alertingsignals are special simple signalsdesigned to attractthe receiver's attentionbeforethe mainsignalis sent; examples of alerting signals are foundforboth vision (Fleishman 1988) and hearing(Wiley and Richards 1982). Signal-processingmechanismsmay set limitsto signal receptionand, hence, bias the directionof the evolutionof signals and signalreceptors.Response rate is a good example. In vision, ifa flickeror motionis too rapid,thenthe signal is degraded or missed entirely(Lythgoe 1979); this is called flickerfusion. Flicker fusioncan set an upper limitto the rate of displays, althoughthis limitmay not be reached in some species (e.g., butterflies; Magnus 1958). In fishthereis a wide rangein the maximumfrequencyat whichflickeris no longerresolved,from14 to 67 Hz (Lythgoe 1979). In general,faster-moving species have higherflicker-fusion frequencies,which are needed in order to track and visually investigatemore rapidlymoving objects (and backgrounds): anotherexample of the correlation between sensory propertiesand behavior. Flicker-fusionfrequencydecreases with lightintensity(Lythgoe 1979), so one can predictthat motion associated with visual displays at low lightlevels should, on the average, be slower than thatof visual displays at higherlightintensities.Flicker-fusion frequencyis also at least in vertebrates;blue-sensitivecones tend to have wavelength-dependent, This content downloaded from 152.14.136.96 on Mon, 3 Jun 2013 09:12:05 AM All use subject to JSTOR Terms and Conditions SIGNALS AND THE DIRECTION OF EVOLUTION S147 lower flicker-fusion rates thanmiddle-or long-wavelength cones. Thus, blue and green should be used in rapid visual displays less oftenthan yellow and red, especiallyat lower lightintensities.However, iftherate of motionis an important partof motion,theninterference colors, whichchangetotalreflectanceand color as a functionof visual angle, are superiorto othercolors, particularlyat short wavelengths.In hearing,the temporalresponse is fasterat higherfrequencies, leading to a trade-offbetween spectral and temporalinformationin bird songs (Okanoya and Dooling 1988). If environmental conditionsfavorsongs withparticular frequencies,thenthese conditionswill affectwhat kindsof information can be transmittedand how rapidly.Anotherexample is visual acuity, the smallest visual angle that can be resolved (Goldsmith 1990). Because visual acuity decreases with decreasing light intensity,color patternswith larger spots and patches are predictedto be morecommonin signalssentat lower lightintensities. Visual acuityalso sets the lower limitto spot size thatis used in signalingat any lightintensity. AvoidingPredation All of the foregoingfactorscan be used to predictthe directionof the joint evolutionof signal properties,receptorproperties,behavior, and microhabitats of species. But thisignoresthe factthatsignalsbetweenconspecificscan be used by predators,so many signals may reflectan evolutionarycompromisebetween signalingto conspecificsand avoidance of signalingto potentialpredators(Endler 1978, 1983, 1986, 1988, 1991; Tuttleand Ryan 1982;Ryan 1985, 1988a). Predicting where the compromiseshould fall is much more difficult than simplypredicting the best signal. However, thereare at least fiveways to reduce the need for a compromisedsignal. First, choose times and places that serve to increase the distance at which predatorsdetect and recognize prey while at the same time not degradingthe signalto conspecifics.For example, signalersthatsignalfromshade can perceive otheranimals (includingpredators)at a greaterdistance thanthose thatsignal in the sun (Helfman 1981). There are two possible reasons for this: (1) the sunlit viewer(predator)has a raised contrast-perception threshold,whichmakes it more difficult to detect contrastat the relativelylow lightlevels in the shade, and (2) veilinglightis greaterin sunlitwater(or foggyor dustyair) thanin shaded water (or air). Shade signalingprobablyworks in fish(Helfman 1981) and could work in other aquatic or terrestrialanimals. In aquatic environmentsthe distance at which predatorsfirstdetect prey can also be decreased by prey's choosing to signal adjacent conspecificsin water thatattenuatesthe signal rapidly.In water or on land thereis an additionaladvantageforaposematic animals to signalfrom a greaterdistance. This allows more timeforan approachingpredatorto remember the noxious propertiesof the signalerand to decide correctlynot to attack (Guilford1986, 1990). Second, aggregatewithotherconspecifics.Aggregationmakes any single signaler less vulnerableto predation(Endler 1986), but it leads to a trade-offwith interference among adjacent signalers(Brush and Narins 1989; Narins 1992). For aposematic prey, aggregationmay also serve to enhance the signal or its rein- This content downloaded from 152.14.136.96 on Mon, 3 Jun 2013 09:12:05 AM All use subject to JSTOR Terms and Conditions S148 THE AMERICAN NATURALIST in forcementin the predator'sbrain(Guilford1990). Third,displayintermittently orderto reduce the probabilitythata predatorwill be presentto detector locate also reducestheprobabilitythata conspecific the signal.Displayingintermittently will receive the signal, unless calling and activitytimes are synchronizedor at least cued by specificenvironmentalfactors.Fourth,use signalsthatdegrade or attenuaterapidlywith distance so that they will be detectable only over short distances. In this way, predators are less likely to sense the signals than the sensorymodes than conspecifics(Wiley and Richards 1982). Fifth,use different by usingdifferpredators(privatechannels), or use the same channelsdifferently ent "tuning" properties,as in guppies. Because guppyvision differsfromthatof theirpredators,theircolor and brightnesscontrastcan be greaterto otherguppies than to the predators(Endler 1991). All of these methodsreduce the need for a given signalto be a compromisebetweenconspecificsignalsand predatoravoidance, but most of these possibilitiesare virtuallyunexplored. CONCLUSIONS Sensory systems,the signals thattheyreceive, signalingbehavior,and microhabitatselection should evolve togetherbecause each induces naturalselection (includingsexual selection; Endler 1986) actingon the other(fig.2). The physics of the environmentand thegeneralbiophysicalpropertiesof signalsand receivers can be used to predictthe directionof evolutionof these suites of traits.There functionsof the same signal should oftenbe conflicting requirementsfordifferent and sensorysystem,as betweenpredationand sexual selection.We mustbe very carefulnot to assume thatonly one functionor one process affectsa traitor our predictionsmay fail. Willson and Whelan's (1990) even-handedapproach to a discussion of fruitcolor is an excellent example of a careful considerationof multiplefunctionsand constraints.Conflictsarisingfrommultiplefunctionsand constraintscan be at least partiallyresolvedby the use of different senses, different sensory characteristics,appropriatebehavior, and appropriatemicrohabitat vision in choice (Endler 1978, 1983, 1991; Alberts 1989, 1992) or even different different partsof the same eye (Bernardand Remington1991). Multiplefunctions and constraintshave not been explored in any detail in any system. From the foregoingdiscussion, it is clear thatbehavioris intimatelyrelatedto successful communication;it is not sufficient to say thatcommunicationmerely throughthe environmentand depends on signals' being successfullytransmitted received by the receptors. Specific behavior is requiredin order to choose the with times, seasons, and microhabitatsthat transmitthe signal most efficiently the least degradationand attenuation,have the least ambientnoise, and have minimalriskof predation.Thus, both breedingbehaviorand microhabitatchoice are likelyto coevolve withsensorysystemsand signals(fig.2). The resultshould be a correlationbetween sensorysystems,signals,signalingbehavior,and microhabitatchoice. Because the sensorysystemis underdirectselection fromfuncand predatorescape, tionsnotdirectlyrelatedto matechoice, such as foodfinding the directionof evolutionof mate choice and otherbehavior can be affectedby sensorybiology. This content downloaded from 152.14.136.96 on Mon, 3 Jun 2013 09:12:05 AM All use subject to JSTOR Terms and Conditions SIGNALS AND THE DIRECTION OF EVOLUTION S149 The evolutionarybias of sensorydriveis unlikelyto be geographicallyuniform, and speciation. a factthathas strongimplicationsforgeographicaldifferentiation Female choice criteriamay be geographicallyvariable (Houde and Endler 1990) and can evolve, sometimesrapidly,under sexual selection (Fisher 1930; Lande 1981; Kirkpatrick1982; Pomiankowski1988). Even if femalechoice were unimportant,microhabitats,times,and seasons of communicationwould varyamong populationsand species. Because signaltransmissionconditionsand background noise also vary with microhabitats,times, and seasons, divergenceof behavior and signaland receptorpropertiesamongpopulationsand species is likely.Variation in intraspecificsignals may not only maintainbut also cause differences among populations. Speciation and furtherdivergencecan resultif populations directions(Lande 1981, 1982; Kirkpatrick1982; and species evolve in different Lande and Kirkpatrick1988). Furtherdivergenceis enhancedthroughdivergence in microhabitatchoice. For example, for poikilotherms,especially small ones, such as insects, variationin microhabitatis associated with strongdifferences in temperature,humidity,thermalbalance, posture,and otherthermallyrelated behavior, favoringdifferencesin size and shape. Microhabitatvariationis also associated with differencesin water balance, parasites and parasitoids,success and durationof egg and larval development,fecundity,reproduction,orientation, and orientationcues (Willmer 1982). This should resultin suites of apparently unrelatedtraitsas disparateas senses, signals,behavior,and physiologyevolving in concert, thus producingpopulations and, eventually,species that differin multiplecharacteristicways. In principle,these sets of apparentlyunrelatedtraits should be predictable,but only if one suite of traitsis consideredin the context and the othersuites. Thus a considerationof thejoint of the microenvironment evolution of signals, sensory systems,and theirassociated behavior may yield new insightsinto the mechanismsof the evolutionand divergenceof species. ACKNOWLEDGMENTS I am most gratefulfor helpfulcommentson the manuscriptfrom,and discussions with, A. Basolo, G. Bernard, T. Guilford,A. Houde, C. Langtimm, K. Long, G. Rosenqvist, P. Ross, V. Rush, M. Ryan, and C. Sandoval. I am also happy to acknowledge financialsupportfromthe Hasselblad Foundation, the National Geographic Society, and the National Science Foundation. LITERATURE CITED Aho, A.-C., K. Donner, C. Hyden, L. 0. Larsen, and T. Reuter. 1988. Low retinalnoise in animals withlow body temperatureallows highvisual sensitivity.Nature (London) 334:348-350. Alberts, A. C. 1989. Ultravioletvisual sensitivityin desert iguanas: implicationsfor pheromone detection.Animal Behaviour 38:129-137. . 1992. Constraintson the designof chemicalcommunicationsystemsin terrestrial vertebrates. AmericanNaturalist139:S62-S89. Alexander, R. A. 1962. Evolutionary change in cricket acoustical communication. Evolution 16:443-467. Anderson,M. E., and R. N. Conner. 1985. Northerncardinalsong in threeforesthabitatsin eastern Texas. Wilson Bulletin97:436-449. This content downloaded from 152.14.136.96 on Mon, 3 Jun 2013 09:12:05 AM All use subject to JSTOR Terms and Conditions S 150 THE AMERICAN NATURALIST Archer,S. N., J. A. Endler, J. N. Lythgoe,and J. C. Partridge.1987. Visual pigmentpolymorphism in the guppyPoecilia reticulata.Vision Research 28:1243-1252. Basolo, A. L. 1990. Female preferencepredatesthe evolutionof the sword in swordtailfish.Science (Washington,D.C.) 250:808-810. Bernard,G. D., and C. L. Remington.1991. Color vision in Lycaena butterflies:spectraltuningof receptorarrays in relationto behavioral ecology. Proceedingsof the National Academy of Sciences of the USA 88:2783-2787. Blakemore, C., ed. 1990. Vision: coding and efficiency.CambridgeUniversityPress, Cambridge. Bowman, R. I. 1979. Adaptivemorphologyof song dialectsin Darwin's finches.JournalfurOrnithologie 120:353-389. Brenowitz, E. A. 1986. Environmentalinfluenceson acoustic and electric animal communication. Brain Behavior and Evolution 28:32-42. Brush, J. S., and P. M. Narins. 1989. Chorus dynamics of a neotropicalamphibianassemblage: comparisonof computersimulationand naturalbehaviour. AnimalBehaviour 37:33-44. Capranica, R. R., and A. J. M. Moffat.1983. Neurobehavioralcorrelatesof sound communication in anurans. Pages 701-730 in J.-P. Ewert, R. R. Capranica, and D. J. Ingle, eds. Advances in vertebrateneuroethology.NATO ASI series A: life sciences. Vol. 56. Plenum, New York. Christy,J. H. 1988. Pillarfunctionin thefiddlercrab Uca beebei. II. Competitivecourtshipsignalling. Ethology78:113-128. Cohen, J. A. 1984. Sexual selectionand the psychophysicsof femalechoice. Zeitschrift furTierpsychologie 64:1-8. Dooling, R. J. 1982. Auditoryperceptionin birds. Pages 96-130 in D. Kroodsma, E. H. Miller,and H. Ouellet, eds. Acoustic communicationin birds. Vol. 1. Academic Press, New York. Endler, J. A. 1978. A predator's view of animal color patterns.EvolutionaryBiology 11:319-364. 1980. Natural selection on color patternsin Poecilia reticulata.Evolution 34:76-91. 1983. Naturaland sexual selectionon color patternsin poeciliidfishes.EnvironmentalBiology of Fishes 9:173-190. 1984. Progressive backgroundmatchingin moths, and a quantitativemeasure of crypsis. Biological Journalof the Linnean Society 22:187-231. 1986. Defense against predation. Pages 109-134 in M. E. Feder and G. V. Lauder, eds. Predator-preyrelationships,perspectives and approaches fromthe study of lower vertebrates. Universityof Chicago Press, Chicago. 1987. Predation,lightintensity,and courtshipbehaviourin Poecilia reticulata.AnimalBehaviour 35:1376-1385. 1988. Frequency-dependentpredation,crypsis, and aposematic coloration. Philosophical Transactionsof the Royal Society of London B, Biological Sciences 319:505-523. 1989. Conceptual and other problems in speciation. Pages 625-648 in D. Otte and J. A. Endler, eds. Speciation and its consequences. Sinauer, Sunderland,Mass. 1990. On the measurementand classificationof color in studies of animal color patterns. Biological Journalof the Linnean Society 41:315-352. 1991. Variation in the appearance of guppy color patternsto guppies and theirpredators visual conditions.Vision Research 31:587-608. underdifferent Endler,J. A., and T. McLellan. 1988. The processes of evolution:towardsa newersynthesis.Annual Review of Ecology and Systematics19:395-421. 1983.Neurologicalanalysisof theinnatereleasing Ewert,J.-P.,H. Burghagen,and E. Schurg-Pfeiffer. mechanismforprey-catching behaviorin toads. Pages 413-475 in J.-P. Ewert,R. R. Capranica, and D. J. Ingle, eds. Advances in vertebrateneuroethology.NATO ASI series A: life sciences. Vol. 56. Plenum, New York. Ewing, A. W. 1983. Functionalaspects ofDrosophila courtship.Biological Reviews of the Cambridge Philosophical Society 58:275-292. Fisher, R. A. 1930. The geneticaltheoryof naturalselection. Clarendon,Oxford. Fleishman, L. J. 1986. Motion detectionin the presence and absence of backgroundmotionin an Anolis lizard. Journalof ComparativePhysiologyA, Sensory,Neural, and Behavioral Physiology 159:711-720. This content downloaded from 152.14.136.96 on Mon, 3 Jun 2013 09:12:05 AM All use subject to JSTOR Terms and Conditions SIGNALS AND THE DIRECTION OF EVOLUTION S151 . 1988. Sensory influences oni physical design of a visual display. Animal Behaviour 36:1420-1424. . 1992. The influenceof the sensory system and the environmenton motion patterns in the visual displays of anoline lizards and other vertebrates.American Naturalist 139: S36-S61. Gish, S. L., and E. S. Morton. 1981. Structuraladaptations to local habitatacoustics in Carolina wren songs. ZeitschriftfurTierpsychologie56:74-84. Goldsmith,T. H. 1990. Optimization,constraint,and historyin the evolutionof the eyes. Quarterly Review of Biology 65:281-322. Grafen,A. 1990. Sexual selection unhandicappedby the Fisher process. Journalof TheoreticalBiology 144:473-516. Greenfield,M. D. 1988. Interspecificacoustic interactionsamong katydids Neoconocephalus: inhibition-induced shiftsin diel periodicity.Animal Behaviour 36:684-695. Guilford,T. 1986. How do warningcolors work? conspicuousness may reduce recognitionerrorsin experienced predators.Animal Behaviour 34:286-288. 1990. The evolution of aposematism. Pages 23-61 in D. L. Evans and J. 0. Schmidt,eds. Insect defenses: adaptive mechanismsand strategiesof preyand predators.State University of New York, Albany. Guilford,T., and M. S. Dawkins. 1991. Receiver psychologyand the evolution of animal signals. Behaviour 42:1-14. Aninmal Guthrie,D. M. 1983. Visual central process in fishbehavior. Pages 381-412 in J.-P. Ewert, R. R. Capranica, and D. J. Ingle, eds. Advances in vertebrateneuroethology.NATO ASI series A: life sciences. Vol. 56. Plenum,New York. Halliday, T. R., and P. J. B. Slater. 1983. Communication.W. H. Freeman, New York. Helfman,G. S. 1981. The advantage to fishesof hoveringin shade. Copeia 1981:392-400. Henwood, K., and A. Fabrick. 1979. A quantitativeanalysis of the dawn chorus: temporalselection forcommunicatoryoptimization.AmericanNaturalist114:260-274. Heth, G., E. Frankenberg,and E. Nevo. 1986. Adaptive optimalsound forvocal communicationin tunnelsof a subterraneanmammal(Spalax ehrenbergi).Experientia(Basel) 42:1287-1289. Houde, A. E. 1987. Mate choice based upon naturallyoccurringcolor-patternvariationin a guppy population. Evolution 41:1-10. 1988a. The effectsof femalechoice and male-malecompetitionon the matingsuccess of male guppies. Animal Behaviour 36:888-896. 1988b.Geneticdifferencein femalechoice betweentwoguppypopulations.AnimalBehaviour 36:510-516. Houde, A. E., and J. A. Endler. 1990. Correlated evolution of female mating preferences and male color patternsin the guppy,Poecilia reticulata.Science (Washington,D.C.) 248: 1405-1408. Hunter,M. L., Jr.,and J. R. Krebs. 1979. Geographicalvariationin the song of the greattit(Pariis major) in relationto ecological factors.Journalof Animal Ecology 48:759-785. Hurvich,L. 1981. Color vision. Sinauer, Sunderland,Mass. Jacobs, G. H. 1981. Comparativecolor vision. Academic Press, New York. Jones,M. D. R. 1966. The acoustic behaviorofthebush cricketPhoildopteragriseoaptera.I. Alternation, synchronismand rivalrybetween males. Journalof ExperimentalBiology 45:31-44. Kirkpatrick,M. 1982. Sexual selection and the evolutionof femalechoice. Evolution 36:1-12. Kodric-Brown,A. 1985. Female preferenceand sexual selection for male coloration in the guppy (Poecilia reticulata). Behavioral Ecology and Sociobiology 17:199-205. Lande, R. 1981. Models of speciationby sexual selectionon polygeniccharacters.Proceedingsof the National Academy of Sciences of the USA 78:3721-3725. 1982. Rapid origin of sexual isolation and character divergence in a cline. Evolution 36:213-223. Lande, R., and M. Kirkpatrick.1988. Ecological speciationby sexual selection.JournalofTheoretical Biology 133:85-98. Laughlin, S. 1981. Neural principlesin the peripheralsystemsof invertebrates.Pages 135-280 in H. Autrum,ed. Comparative physiologyand evolution in invertebrates,B: invertebrate This content downloaded from 152.14.136.96 on Mon, 3 Jun 2013 09:12:05 AM All use subject to JSTOR Terms and Conditions S152 THE AMERICAN NATURALIST visual centers and behaviour I. Handbook of sensory physiology.Vol. VII/8B. Springer, New York. Levine, J. S., and E. F. MacNichol, Jr. 1979. Visual pigmentsin teleost fishes: effectsof habitat, microhabitat,and behavior on visual systemevolution. Sensory Processes 3:95-131. Long, K. D., and A. E. Houde. 1989. Orange spots as a visual cue forfemale mate choice in the guppy (Poecilia reticulata). Ethology82:316-324. Lythgoe,J. N. 1979. The ecology of vision. OxfordUniversityPress, Oxford. Lythgoe, J. N., and J. C. Partridge.1991. The modellingof optimalvisual pigmentsof dichromatic teleosts in green coastal waters. Vision Research 31:361-371. Magnus, D. 1958. ExperimentelleUntersuchungenzur Bionomie und Ethologie des Aisermantels Argynnispaphia Girard (Lepidoptera: Nymphaidae). ZeitschriftfuirTierpsychologie 15: 397-426. Markow, T. A., and E. C. Toolson. 1990. Temperatureeffectson epicuticularhydrocarbonsand sexual isolationin Drosophila mojavensis. Pages 315-331 in J. S. F. Barker,W. T. Starmer, and R. J. Maclntyre,eds. Ecological and evolutionarygeneticsof Drosophila. Plenum,New York. rules. Pages Morton, E. S. 1982. Grading, discreteness, redundancy, and motivation-structural 183-212 in D. E. Kroodsma and E. K. Miller,eds. Acoustic communicationin birds. Vol. 1. Academic Press, New York. 1990. Habitat segregationby sex in the hooded warbler:experimentson proximatecausation and discussion of its evolution. AmericanNaturalist135:319-333. Narins,P. M. 1992. Evolutionof anuranchorusbehavior:neuraland behavioralconstraints.American Naturalist139:S90-S104. Okanoya, K., and R. J. Dooling. 1988. Hearingin the swamp sparrow,Melospiza georgiana, and the song sparrow,Melospiza melodia. Animal Behaviour 36:726-732. Pomiankowski,A. 1988. The evolutionof female mate preferencesformale geneticquality. Oxford Surveys in EvolutionaryBiology 6:136-184. Reimchen, T. E. 1989. Loss of nuptial color in threespinesticklebacks(Gasterosteus aculeatus). Evolution 43:450-460. Richards, D. G., and R. H. Wiley. 1980. Reverberationsand amplitudefluctuationsin the propagation of sound in a forest: implicationsfor animal communication.American Naturalist 115:381-399. Ryan, M. J. 1985. The tuingarafrog:a studyin sexual selection and communication.Universityof Chicago Press, Chicago. 1988a. Constraintsand patternsin the evolutionof anuran acoustic communication.Pages 637-675 in B. Fritzsch,M. J. Ryan, W. Wilczynski,T. E. Hetherington, and W. Walkowiak, eds. The evolutionof the amphibianauditorysystem.Wiley, New York. 1988b. Energy,calling,and selection. AmericanZoologist 28:885-898. 1990. Sexual selection,sensorysystems,and sensoryexploitation.OxfordSurveys in EvolutionaryBiology. 7:156-195. Ryan, M. J., and E. A. Brenowitz. 1985. The role of body size, phylogeny,and ambientnoise in the evolutionof bird song. AmericanNaturalist126:87-100. Ryan, M. J., and A. Keddy-Hector. 1992. Directionalpatternsof femalemate choice and the role of sensorybiases. American Naturalist139:S4-S35. Ryan, M. J., and A. S. Rand. 1990. The sensorybasis of sexual selection for complex calls in the tulngarafrog,Physalaemus pustulosus (sexual selectionforsensoryexploitation).Evolution 44:305-314. Ryan, M. J., R. B. Cocroft,and W. Wilczynski.1990. The role of environmental selectionin intraspecific divergenceof mate recognitionsignals in the cricketfrog,Acris crepitans. Evolution 44:1869-1872. Sebeok, T. A., ed. 1977. How animals communicate.Indiana UniversityPress, Bloomington. Seliger, H. H., A. B. Lall, J. E. Lloyd, and W. H. Beggley. 1982. The colors of firefly bioluminescence. I. Optimizationmodel. Photochemistry and Photobiology36:673-680. Simmons,L. W. 1988. The callingsong of thefieldcricket,Gryllusbimaculatus(De Geer): constraints This content downloaded from 152.14.136.96 on Mon, 3 Jun 2013 09:12:05 AM All use subject to JSTOR Terms and Conditions SIGNALS AND THE DIRECTION OF EVOLUTION S153 on transmissionand its role in intermalecompetitionand femalechoice. Animal Behaviour 36:380-394. Stone, J. 1983. Parallel processingin the visual system:the classificationof retinalganglioncells and its impact on the neurobiologyof vision. Plenum,New York. in femalepreferencerelatedto geographic Stoner,G., and F. Breden. 1988. Phenotypicdifferentiation variationin male predationriskin the Trinidadguppy(Poecilia reticulata).Behavioral Ecology and Sociobiology 22:285-291. Tuttle,M. D., and M. J. Ryan. 1982.The role of synchronizedcalling,ambientlight,and ambientnoise in anti-bat-predator behaviorof a treefrog.Behavioral Ecology and Sociobiology 11:125-131. Waser, P. M., and C. H. Brown. 1984. Is there a "sound window" for primatecommunication? Behavioral Ecology and Sociobiology 15:73-76. Waser, P. M., and M. S. Waser. 1977. Experimentalstudies of primatevocalization: specializations forlong-distancepropagation.ZeitschriftfurTierpsychologie43:239-263. Wiley, R. H., and D. G. Richards. 1982. Adaptationsfor acoustic communicationin birds: sound transmissionand signaldetection.Pages 132-181 in D. E. Kroodsma and E. K. Miller,eds. Acoustic communicationin birds. Vol. 1. Academic Press, New York. Willmer,P. G. 1982. Microclimateand the environmentalphysiologyof insects. Advances in Insect Physiology16:1-57. Willson,M. F., and C. J. Whelan. 1990. The evolutionoffruitcolor in fleshy-fruited plants.American Natutalist 136:790-809. This content downloaded from 152.14.136.96 on Mon, 3 Jun 2013 09:12:05 AM All use subject to JSTOR Terms and Conditions