1: \documentclass[11pt,a4paper]{article}

2: \usepackage{fancyhdr}

3: \usepackage{palatino}

4: \usepackage{amssymb}

5: \usepackage{epsfig}

6: \usepackage{vmargin}

7: \usepackage{natbib}

8: 

9: % Pagestyle

10: \pagestyle{fancy}

11: 

12: % \headrulewidth is width of line under head.

13: \renewcommand{\headrulewidth}{0.4pt}

14: 

15: % \footrulewidth is width of line above foot. 

16: % \renewcommand{\footrulewidth}{\headrulewidth}

17: 

18: % What the head and foot should consist of.

19: \fancyhf{}

20: \fancyhead[L]{Cartwright, Gonz\'alez, \& Piro}

21: \fancyfoot[R]{\bfseries\thepage}

22: 

23: %\renewcommand{\baselinestretch}{1.35}

24: 

25: \renewcommand{\refname}{}

26: \makeatletter \renewcommand{\@biblabel}[1]{#1.} \makeatother

27: \renewcommand\figurename{{\sffamily\bfseries Figure}}

28: 

29: % Arrangement of figures on page

30: \renewcommand{\topfraction}{0.95}

31: \renewcommand{\bottomfraction}{0.95}

32: \renewcommand{\textfraction}{0.05}

33: \renewcommand{\floatpagefraction}{0.1}

34: \setcounter{topnumber}{1}

35: \setcounter{bottomnumber}{3}

36: \setcounter{totalnumber}{3}

37: 

38: \begin{document}

39: 

40: \begin{bfseries}

41: \noindent\Large{Pitch perception: A dynamical-systems perspective}

42: \vspace{0.5cm}

43: 

44: \noindent\small{Julyan H. E. Cartwright*, Diego L. Gonz\'alez\dag,

45:  \& Oreste Piro\ddag}

46: \end{bfseries}

47: 

48: \begin{itshape}

49: \noindent\footnotesize

50: *~Laboratorio de Estudios Cristalogr\'aficos, CSIC, E-18071 Granada, Spain. \\

51: E-mail julyan@lec.ugr.es,  Web http://lec.ugr.es/$\sim$julyan \\

52: \dag~Istituto Lamel, CNR, I-40129 Bologna, Italy. \\

53: E-mail gonzalez@lamel.bo.cnr.it \\

54: \ddag~Institut Mediterrani d'Estudis Avan\c{c}ats, CSIC--UIB, 

55: E-07071 Palma de Mallorca, Spain. \\

56: E-mail piro@imedea.uib.es, Web http://www.imedea.uib.es/$\sim$piro \\

57: \end{itshape}

58: 

59: \normalsize

60: 

61: Published in: Proc. Nat. Acad. Sci. USA {\bf 98}, 4855-4859, 2001.

62: 

63: \vspace{0.5cm}

64: 

65: %{\bf Classification:} Physical Sciences (Physics) \& Biological Sciences

66: %(Neurobiology)

67: 

68: %{\bf Manuscript information:} 13 pages; 4 figures

69: 

70: %{\bf Word and character counts:}  Abstract 212 words; paper 40\,798 characters.

71: 

72: %\newpage

73: 

74: \begin{bfseries}

75: \noindent\small

76: Two and a half millennia ago Pythagoras initiated the scientific study of the

77: pitch of sounds; yet our understanding of the mechanisms of pitch perception

78: remains incomplete. Physical models of pitch perception try to explain from

79: elementary principles why certain physical characteristics of the stimulus lead

80: to particular pitch sensations. There are two broad categories of

81: pitch-perception models: place or spectral models consider that pitch is mainly

82: related to the Fourier spectrum of the stimulus, whereas for periodicity or

83: temporal models its characteristics in the time domain are more important.

84: Current models from either class are usually computationally intensive,

85: implementing a series of steps more or less supported by auditory physiology.

86: However, the brain has to analyse and react in real time to an enormous amount

87: of information from the ear and other senses. How is all this information

88: efficiently represented and processed in the nervous system? A proposal of

89: nonlinear and complex systems research is that dynamical attractors may form

90: the basis of neural information processing. Because the auditory system is a 

91: complex and highly nonlinear dynamical system it is natural to suppose that

92: dynamical attractors may carry perceptual and functional meaning. Here we show

93: that this idea, scarcely developed in current pitch models, can be successfully

94: applied to pitch perception.

95: \end{bfseries}

96: 

97: The pitch of a sound is where we perceive it to lie on a musical scale. For a

98: pure tone with a single frequency component, pitch rises monotonically with

99: frequency. However, more complex signals also elicit a pitch

100: sensation. Some instances are presented in Fig.~\ref{stimuli}. These are sounds

101: produced by the nonlinear interaction of two or more periodic sources, by

102: amplitude or frequency modulation. All such stimuli, which may be termed

103: complex tones, produce a definite pitch sensation, and all of them exhibit a

104: certain spectral periodicity. Many natural sounds have this quality,

105: including vowel sounds in human speech and vocalizations of many other

106: animals. Evidence for the importance of spectral periodicity in sound

107: processing by humans is that noisy stimuli exhibiting this property also

108: elicit a pitch sensation. An example is repetition pitch: the pitch of 

109: ripple noise \cite{yost}, which arises naturally when the sound from a noisy

110: source interacts with a delayed version of itself, produced, for example, by

111: a single or multiple echo. It is clear that an efficient mechanism for the

112: analysis and recognition of complex tones represents an evolutionary advantage

113: for an organism. In this light, the pitch percept may be seen as an effective

114: one-parameter categorization of sounds possessing some spectral periodicity

115: \cite{bregman,hartmann,roberts,moore}. 

116: 

117: \begin{figure}[p]

118: \begin{center}

119: \epsfig{file=nat1.eps,width=0.8\textwidth}

120: \end{center}

121: \caption{\label{stimuli}

122: Stimuli: waveforms, Fourier spectra, and pitches.

123: (a) 1 kHz pure tone; the pitch coincides with the frequency $\omega_0$.

124: (b) Complex tone formed by 200 Hz fundamental plus overtones; the pitch is at

125: the frequency of the fundamental $\omega_0$.

126: (c) After high-pass filtering of the previous tone to remove

127: the fundamental and the first few overtones, the pitch $\omega_0$ remains at the

128: frequency of the missing fundamental (dotted).

129: (d) The result of frequency modulation of a 1 kHz pure tone carrier

130: by a 200 Hz pure tone modulant.

131: (e) Complex tone produced by amplitude modulation of a 1 kHz pure tone carrier

132: by a 200 Hz pure tone modulant; the pitch coincides with the difference 

133: combination tone $\omega_0$.

134: (f) Result of shifting the partials of the previous tone in frequency by 

135: $\Delta\omega=90$ Hz; the pitch shifts by $\Delta\omega_0\approx 20$ Hz,

136: although the difference combination tone does not.

137: (g) Schematic diagram of the frequency line details (above the line)

138: the pitch shift behaviour of (f) and (below the line) the three-frequency 

139: resonance we propose to explain it.

140: }

141: \end{figure}

142: 

143: \begin{center}\bf Virtual Pitch\end{center}

144: 

145: For a harmonic stimulus like Fig.~\ref{stimuli}b (a periodic signal), 

146: there is a natural physical 

147: solution to the problem of encoding it with a single parameter: take the

148: fundamental component of the stimulus as the pitch and all other components are

149: naturally recorded as the higher harmonics of the fundamental. This is what 

150: nature does. However, a harmonic stimulus like Fig.~\ref{stimuli}c, which is 

151: high-pass filtered so that the fundamental and some of the first higher

152: harmonics are eliminated, nevertheless maintains its pitch at the frequency of

153: the absent fundamental. The stimulus (Fig.~\ref{stimuli}e) obtained by

154: amplitude modulation of a sinusoidal carrier of 1 kHz by a sinusoidal modulant

155: of 200 Hz is also of this type. As the carrier and modulant are rationally

156: related, the stimulus is harmonic; the partials are integer multiples of the

157: absent fundamental $\omega_0=200$ Hz. The perception of pitch for this kind of

158: stimulus is known as the problem of the missing fundamental, virtual pitch, or

159: residue perception \cite{deboer}. The first physical theory  for the phenomenon

160: was proposed by von Helmholtz \cite{helmholtz}, who attributed it to the

161: generation of difference combination tones in the nonlinearities of the ear. A

162: passive nonlinearity fed by two sources with frequencies $\omega_1$ and

163: $\omega_2$ generates combination tones of frequency $\omega_C$ (see the 

164: Appendix for clarification of the concepts from nonlinear dynamics used

165: throughout this paper). For a harmonic complex tone, such as

166: Fig.~\ref{stimuli}e, the difference combination tone

167: $\omega_C=\omega_2-\omega_1$ between two successive partials has the frequency

168: of the missing fundamental $\omega_0$. In a crucial experiment, however,

169: Schouten et al.\ \cite{schouten} demonstrated that the  residue cannot be

170: described by a difference combination tone: if we shift all the partials in

171: frequency by the same amount $\Delta\omega$  (Fig.~\ref{stimuli}f), the

172: difference combination tone remains unchanged. But the perceived pitch shifts,

173: with a linear dependence on $\Delta\omega$.

174: 

175: \begin{center}\bf A Dynamical-Systems Perspective\end{center}

176: 

177: Such a complex tone is no longer harmonic. How does nature encode an inharmonic

178: complex tone into a single pitch? Intuitively, the shifted pseudofundamental

179: depicted in Fig.~\ref{stimuli}g might seem to be a better choice than the

180: unshifted fundamental, which corresponds to the difference combination tone.

181: However, from a mathematical point of view, this is not obvious. The ratios

182: between successive partials of the shifted stimulus are irrational and we

183: cannot represent them as higher harmonics of a nonzero fundamental frequency; 

184: the true fundamental would have frequency zero. Some kind of

185: approximation is needed. The approximation of two arbitrary frequencies,

186: $\omega_1$ and $\omega_2$, by the harmonics of a third, $\omega_R$, is

187: equivalent to the mathematical problem of finding a strongly convergent

188: sequence of pairs of rational numbers with the same denominator that

189: simultaneously approximates the two frequency ratios, $\omega_1/\omega_R$ and

190: $\omega_2/\omega_R$. If we consider the approximation to only one frequency

191: ratio there exists a general solution given by the continued-fraction algorithm

192: \cite{kinchin}. However, for two frequency ratios a general solution is not

193: known. Some algorithms have been proposed that work for particular values

194: of the frequency ratios or that are weakly convergent \cite{kim1}. We developed

195: an alternative approach \cite{3freq}. The idea is to equate the distances

196: between appropriate harmonics of the pseudofundamental and the pair of

197: frequencies we wish to approximate. In this way the two approximations are

198: equally good or bad. The problem can then be solved by a generalization of the

199: Farey sum. This approach enables the hierarchical

200: classification of a type of dynamical attractors found in systems with three

201: frequencies: three-frequency resonances $[p,q,r]$.  

202: 

203: A classification of three-frequency resonances allows us to propose how nature

204: might encode an inharmonic complex tone into a single pitch percept. 

205: The pitch of a complex tone corresponds to a one-parameter

206: categorization of sounds by a physical frequency whose harmonics are good

207: approximations to the partials of the complex. This physical frequency is

208: naturally generated as a universal response of a nonlinear dynamical system 

209: --- the auditory system, or some specialized subsystem of it --- 

210: under the action of an external force, namely the

211: stimulus. Psychophysical experiments with multicomponent stimuli suggest that

212: the lowest-frequency components are usually dominant in determining residue

213: perception \cite{deboer}. Thus we represent the external force as a first

214: approximation by the two lowest-frequency components of the stimulus. For pitch

215: shift experiments with small frequency detuning $\Delta\omega$, such as those

216: of Schouten et al., the vicinity of these two lowest components

217: $\omega_1=k\omega_0+\Delta\omega$ and $\omega_2=(k+1)\omega_0+\Delta\omega$ to

218: successive multiples of some missing fundamental ensures that $(k+1)/k$ is a

219: good rational approximation to their frequency ratio. Hence we concentrate on a

220: small interval between the frequencies $\omega_1/k$ and $\omega_2/(k+1)$ around

221: the missing fundamental of the nonshifted case. These frequencies correspond to

222: the three-frequency resonances $[0,-1,k]$ and $[-1,0,k+1]$. We suppose that the

223: residue should be associated with the largest three-frequency resonance in this

224: interval: the daughter of these resonances, $[-1,-1,2k+1]$. If this reasoning is

225: correct, the three-frequency resonance formed between the two lowest-frequency

226: components of the complex tone and the response frequency

227: $P=(\omega_1+\omega_2)/(2k+1)$ gives rise to the perceived residue pitch $P$. 

228: 

229: \begin{center}\bf Results\end{center}

230: 

231: As we showed in earlier work \cite{soundsletterprl}, there is good agreement

232: betwen the pitch perceived in experiments and the three-frequency resonance

233: produced by the two lowest-frequency components of the complex tone for

234: intermediate harmonic numbers $3\leq k\leq 8$. For high and low $k$ values

235: there are systematic deviations from these predictions. Such deviations, noted

236: in pitch-perception modelling, are explained by the dominance effect: there is

237: a frequency window of preferred stimulus components, so that not all components

238: are equally important in 

239: determining residue perception \cite{patterson2}. In order to describe these 

240: slope deviations for high and low $k$ values within our approach, we must,

241: instead of taking the lowest-frequency components, use some effective $k$ 

242: that depends on the dominance effect. In this, we also take into account the

243: presence of difference combination tones, which provide some components with

244: $k$s not present in the original stimulus.  In Fig.~\ref{results} we have

245: superimposed the predicted three-frequency resonances, including the dominance

246: effect, on published experimental pitch-shift data

247: \cite{schouten,gerson,patterson}. For stimuli consisting only of high-$k$

248: components, the window of the dominance region is almost empty, and difference

249: combination tones of lower $k$ can become more important than the primary

250: components in determining the pitch of the stimulus. The result of this

251: modification is a saturation of the slopes that correctly describes the

252: experimental data. A saturation of slopes can also be seen in the experimental

253: data for low values of $k$. This effect too can be explained in terms of the

254: dominance region. For a 200 Hz stimulus spacing, the region is situated at

255: about 800 Hz; this implies that stimulus components with harmonic numbers $n$

256: and $n+1$ other than the two lowest ones (i.e., $n>k$) become more important

257: for determining the the three-frequency resonance that provides the residue

258: pitch. Again, incorporating this modification, we can correctly predict

259: the experimental data.

260: 

261: \begin{figure}[p]

262: \begin{center}

263: \epsfig{file=fern3.eps,width=\textwidth}

264: \end{center}

265: \caption{\label{results}

266: Experimental data (red dots) from Gerson \& Goldstein \cite{gerson} (0--800 Hz

267: range) and from Schouten et al.\ \cite{schouten} (1200--2200 Hz range) show 

268: pitch as a function of the lower frequency $f=k\omega_0+\Delta\omega$ of a 

269: complex tone $\{k\omega_0+\Delta\omega, (k+1)\omega_0+\Delta\omega, 

270: (k+2)\omega_0+\Delta\omega, \ldots\}$ with the partials spaced

271: $g=\omega_0=200$  Hz apart. The data of Schouten et al.\ are for

272: three-component tones monotically presented (all of the stimulus entering one

273: ear), and those of Gerson \& Goldstein for four-component tones dichotically

274: presented (part of the stimulus entering one ear and the rest of the stimulus

275: the other, controlateral, ear); the harmonic numbers of the partials present in

276: the stimuli are shown beside the data. The pitch-shift effect we predict from

277: three-frequency resonance,

278: taking into account the dominance region, is shown superimposed on the data as

279: solid lines given by the equations $P=g+(f-n\,g)/(n+1/2)$ (primary lines)

280: $P=g/2+(f-(n+1/2)g)/(2n+2)$ (secondary lines), and $P=g/4+(f-(n-1/4)g)/(4n+1)$ 

281: (tertiary line); the harmonic numbers of the partials used to 

282: calculate the pitch-shift lines are shown enclosed in red squares. For primary

283: lines these harmonic numbers correspond to $n$ and $n+1$, for secondary lines to

284: $2n+1$ and $2n+3$, and for the tertiary line to $4n+1$ and $4n+5$. A red

285: circle, instead of a square, signifies that the component is not physically

286: present in the stimulus, but corresponds to a combination tone. The inset at

287: bottom right corresponds to the slopes of the data averaged  over the distinct

288: experimental values plotted as a function of harmonic number. The blue squares

289: are the data of Gerson \& Goldstein, the red squares are those of Schouten, and

290: lastly, the blue circles are data of Patterson \cite{patterson} for six and

291: twelve-component tones which are averaged over different experimental

292: situations that represent several thousand points. The black diamonds

293: correspond to our theory and show that the data of Gerson \& Goldstein and

294: those of Patterson saturate for different values of $k$ (the experimental

295: conditions were different). 

296: }

297: \end{figure}

298: 

299: But for the more complex case of low-$k$ stimuli, not only quantitative, but 

300: also qualitative, differences arise between the two-lowest-component theory and

301: experiment. The most interesting feature seen in the data of Fig.~\ref{results}

302: is a second series of pitch-shift  lines clustered around the pitch of 100 Hz. 

303: This too can be explained within the framework of our ideas. Recall that for

304: small frequency detuning, the frequency ratio between adjacent stimulus

305: components, $\Delta\omega$, can be approximated  by the quotient of two

306: integers differing by unity: $\omega_2/\omega_1=(n+1)/n$. However, if we relax

307: the small detuning constraint, so that $\Delta\omega$ becomes large, we can

308: move to a case where $\omega_2/\omega_1$ can better be approximated by

309: $(n+2)/(n+1)$. But, by the usual Farey sum operation between rational numbers,

310: we know that there exists between these two regions an interval in which the

311: frequency ratio can be better approximated by $(2n+3)/(2n+1)$. In this

312: interval, then, the main three-frequency resonance is $[-1,-1,4n+4]$, giving a

313: response frequency $P=(\omega_1+\omega_2)/(4n+4)$, which produces a pitch-shift

314: line with slope $1/(2n+2)$ around $\omega_0/2=100$ Hz for the case analysed. Of

315: course, if prefiltering produces a saturation of the slopes of the primary

316: pitch-shift lines, the same should occur for these secondary ones. In

317: Fig.~\ref{results} we show our predictions for the secondary lines taking in

318: account the dominance effect. The agreement, both qualitative and also

319: quantitative, is impressive. Moreover, a small group of data points indicates

320: the existence of a further level of pitch-shift lines clustered around 50 Hz

321: in a region between a primary and a secondary pitch-shift line. We can

322: understand this level in the same way as above, and we plot our

323: prediction for its pitch-shift line in Fig.~\ref{results}. This 

324: hierarchical arrangement of the perception of pitch of complex tones is

325: entirely consistent with the universal devil's staircase structure that

326: dynamical systems theory predicts for the three-frequency resonances in

327: quasiperiodically forced dynamical systems. Further evidence comes from

328: psychophysical experiments with pure tones. These, presented under particular

329: experimental conditions, also elicit a residue sensation. The extremes of the

330: three-frequency staircase correspond to subharmonics of only one external

331: frequency, and thus these are the expected responses when only one stimulus

332: component is present. As the results of Houtgast \cite{houtgast} show, these

333: subharmonics are indeed perceived. 

334: 

335: \begin{center}\bf Discussion\end{center}

336: 

337: A dynamical attractor can be studied by means of time or frequency analysis.

338: Both are common techniques in dynamical-systems analysis, but one is not

339: inherently more fundamental than the other, nor are these the only two tools

340: available. For this reason, and because our reasoning makes no use of a

341: particular physiological implementation, our results cannot be included

342: directly either in the spectral \cite{cohen2} or the temporal \cite{meddis}

343: classes of models of pitch perception. What we have proposed is not a model,

344: but a mathematical basis for the perception of pitch that uses the universality

345: of responses of dynamical systems to address the question of why the auditory

346: system should behave as it does when confronted by stimuli consisting of

347: complex tones. Not all pitch perception phenomena are explicable in terms of

348: universality; nor should they be, since some will depend on the specific

349: details of the neural circuitry. However, this is a powerful way of approaching

350: the problem that is capable of explaining many experimental data considered

351: difficult to understand. Future pitch models can surely incorporate these

352: results in their frameworks. Spectral models \cite{cohen2} can use these ideas

353: since they make consistent use of different kinds of harmonic templates, and

354: three-frequency resonances offer in a natural way optimized candidates for the

355: base frequency of such templates without the need to include stochastic terms.

356: Temporal models \cite{meddis} can apply these results as they need some kind of

357: locking of neural spiking to the fine structure of the stimulus, and

358: three-frequency resonances are the natural extension of phase locking to the

359: more complicated case of quasiperiodic forcing that is typically related to the

360: perception of complex tones. A dynamical-systems viewpoint can then integrate

361: spectral and temporal hypotheses into a coherent unified approach to pitch

362: perception incorporating both sets of ideas.

363: 

364: We have shown that universal properties of dynamical responses in nonlinear 

365: systems are reflected in the pitch perception of complex tones. In previous

366: work \cite{soundsletterprl}, we argued that a dynamical-systems approach backs

367: up experimental evidence for subcortical pitch processing in humans 

368: \cite{pantev}. The experimental evidence is not conclusive, as studies with

369: monkeys have found that raw spectral information is present in the primary

370: auditory cortex \cite{fishman}. However, whether this processing occurs in, 

371: or before, the auditory cortex, the dynamical mechanism we envisage greatly

372: facilitates processing of information into a single percept. Pitch processing

373: may then prove to be an example in which universality in nonlinear

374: dynamics can help to explain complex experimental results in biology. The

375: auditory system possesses an astonishing capability for processing 

376: pitch-related information in real time; what we have demonstrated here is how, 

377: at a fundamental level, this can be so.

378: 

379: \begin{center}\bf Acknowledgements\end{center}

380: We should like to thank Fernando Acosta for his help in the preparation of 

381: Fig.~\ref{results}. JHEC acknowledges the financial support of the Spanish

382: CSIC, and Plan Nacional del Espacio contract ESP98-1347. OP acknowledges the

383: Spanish Ministerio de Ciencia y Tecnologia, Proyecto CONOCE, contract

384: BFM2000-1108. DLG conceived the idea, and together with JHEC and OP carried 

385: out the research; JHEC and DLG cowrote the paper.

386: 

387: \begin{center}\bf References\end{center}

388: 

389: \bibliographystyle{nature}

390: \bibliography{database}

391: 

392: \newpage

393: \vspace*{0.5cm}

394: \begin{center}\bf Appendix: Universality in Nonlinear Systems\end{center}

395: \vspace{1cm}

396: Nonlinear systems exhibit universal responses under external forcing:

397: \vspace{1cm}

398: \subsubsection*{Harmonics from periodically forced passive nonlinearities}

399: \begin{center}

400: \epsfig{file=figbox1.eps,height=0.85cm}

401: \end{center}

402: A single frequency periodically forcing a passive (sometimes termed static) 

403: nonlinearity generates higher harmonics (overtones)

404: $2\omega_1, 3\omega_1, \ldots$

405: of a fundamental $\omega_1$, given by

406: $p\omega_1+\omega_{H}=0$ with $p$ integer. This is seen in acoustics

407: as harmonic distortion.

408: \vspace{1cm}

409: \subsubsection*{Combination tones from quasiperiodically forced passive 

410: nonlinearities}

411: \begin{center}

412: \epsfig{file=figbox2.eps,height=1cm}

413: \end{center}

414: A passive nonlinearity forced quasiperiodically by two sources generates

415: combination tones $\omega_1-\omega_2, \omega_1+\omega_2, \ldots$, which are 

416: solutions of the equation

417: $p\omega_1+q\omega_2+\omega_{C}=0$ where $p$ and $q$ are integers. They are

418: found as distortion products in acoustics. 

419: \vspace{1cm}

420: \subsubsection*{Subharmonics, or two-frequency resonances from periodically 

421: forced dynamical systems}

422: \begin{center}

423: \epsfig{file=figbox3.eps,height=0.85cm}

424: \end{center}

425: With a periodically forced active nonlinearity --- a dynamical system --- more

426: complex subharmonic responses $\omega_1/r, 2\omega_1/r, \ldots, (r-1)\omega_1/r$

427: known as mode lockings or two-frequency

428: resonances are generated. These are given by $p\omega_1+r\omega_{2R}=0$ when

429: $p$ and $r$ are integers. 

430: As some parameter is varied, different resonances are found that remain

431: stable over an interval. A classical representation of this, known as

432: the devil's staircase, is shown in Fig.~\ref{devils_staircase}.

433: 

434: \begin{figure}[h]

435: \begin{center}

436: \epsfig{file=figbox2fr.eps,height=4cm}\end{center}

437: \caption{\label{devils_staircase}

438: Two-frequency devil's staircase.

439: The rotation number, the frequency 

440: ratio $\rho=-p/r=\omega_{2R}/\omega_1$, is

441: plotted against the period of the external force.

442: }

443: \end{figure}

444: 

445: We see that the resonances are hierarchically arranged. The local ordering can

446: be described by the Farey sum: 

447: If two rational numbers $a/c$ and $b/d$ satisfy $|ad-bc|=1$ we say that they

448: are unimodular or adjacents and we can find between them a unique rational with

449: minimal denominator. This rational is called the mediant and can be expressed

450: as a Farey sum operation $a/c\oplus b/d=(a+b)/(c+d)$. The resonance

451: characterized by the mediant is the widest between those represented by the

452: adjacents \cite{oreste}.

453: \vspace{1cm}

454: \subsubsection*{Three-frequency resonances from quasiperiodically 

455: forced dynamical systems}

456: \begin{center}

457: \epsfig{file=figbox4.eps,height=1cm}

458: \end{center}

459: Quasiperiodically forced dynamical systems show a great variety of

460: qualitative responses that fall into three main categories: there are 

461: periodic attractors, quasiperiodic attractors, and chaotic and nonchaotic 

462: strange attractors. Here we concentrate on the

463: three-frequency resonances produced by two-frequency quasiperiodic attractors 

464: as the natural candidates for modelling the residue \cite{lectnotes}. 

465: Three-frequency resonances are given by the nontrivial

466: solutions of the equation $p\omega_1+q\omega_2+r\omega_{3R}=0$, where $p$, $q$,

467: and $r$ are integers, $\omega_1$ and $\omega_2$ are the forcing frequencies,

468: and $\omega_{3R}$ is the resonant response, and can be written compactly in the

469: form $[p,q,r]$. Combination tones are three-frequency resonances of

470: the restricted class $[p,q,1]$. This is the only type of response possible from

471: a passive nonlinearity, whereas a dynamical system such as a forced oscillator

472: is an active nonlinearity with at least one intrinsic frequency, and can

473: exhibit the full panoply of three-frequency resonances, which include

474: subharmonics of combination tones. Three-frequency resonances obey

475: hierarchical ordering properties very similar to those governing two-frequency

476: resonances in periodically forced systems. In the interval

477: $(\omega_2/p,\omega_1/q)$, we may define a generalized Farey sum between any

478: pair of adjacents as $a_1/c\oplus a_2/d=(a_1+a_2)/(c+d)$. The daughter

479: three-frequency resonance characterized by the generalized mediant is the 

480: widest between its parents characterized by the adjacents \cite{3freq}.

481: Thus three-frequency resonances are ordered very similarly to their

482: counterparts in two-frequency systems, and form their own devil's staircase;

483: Fig.~\ref{3f_devils_staircase}.

484: \vspace{1cm}

485: 

486: \begin{figure}[h]

487: \begin{center}

488: \epsfig{file=figbox3fr.eps,height=4cm}\end{center}

489: \caption{\label{3f_devils_staircase}

490: Three-frequency devil's staircase.

491: Contrarily to the case

492: of periodically driven systems, where plateaux represent periodic

493: solutions, here they represent quasiperiodic solutions

494: (only the third frequency is represented in the ordinate).

495: We have investigated these properties in three different systems: 

496: the quasiperiodic circle map, a system of coupled electronic oscillators 

497: and a set of ordinary nonlinear differential equations, with the same 

498: qualitative results \cite{3freqpaper}, which confirm the theoretical 

499: predictions \cite{3freq}.

500: }

501: \end{figure}

502: 

503: \end{document}

504: