In many situations, the acoustic properties of the bottom are important, and considerable effort has been directed at their elucidation over the whole frequency range of interest to underwater acoustics. At low frequencies, experimental measurement of these properties is particularly difficult and uncertainty exists in the values of some key parameters. The determination of compressional and shear wave attenuation has been especially challenging and it is not surprising that debate has persisted as to the magnitude of this property and its overall dependence on frequency at low frequencies. This paper attempts to examine this question through a review of the relevant literature from rock mechanics and underwater acoustics, with a special emphasis on low-frequency data.

From time to time, strong underwater transient sounds have been detected in New Zealand waters. Much of this activity appears to be biological in origin, and some of it is obviously similar to underwater signals observed in other parts of the world. Other components are by no means so well known and provide further interesting examples of pulses in which the energy is confined to a narrow band of frequencies around 20 cps. These pulses are described.

The application of an M2 nonlinear numerical tidal model to the shelf seas of central New Zealand (∼38,500 km2 area) is described. It has provided a preliminary assessment of tidal and residual currents, bottom stress, energy dissipation, and the stratification index. The existence of a permanent, tidally driven mesoscale eddy (∼75 km diameter) is predicted north of D'Urville Island. Large spatial gradients in bottom stress qualitatively agree with many features of the surficial sediment distribution. A comparison of all available bulk stratification data with the h/u3 stratification index clearly demonstrates the dominance of tidal versus wind mixing over the control of summer stratification. A potential application of the model to fisheries science is suggested through a comparison of the stratification index contour map and some observations of squid fishing vessel locations.

Underwater ambient noise is known to be wind dependent. Several mechanisms have been proposed to explain the nature of the transfer of energy from the wind to the acoustic noise field. Examples include wind and wave turbulence and nonlinear interactions between surface waves. This study examines these wind-related mechanisms at the low end of the acoustic spectrum. Data from a long-term investigation of ocean waves and the associated microseism response recorded ashore have provided evidence helpful in identifying the active processes. It is concluded that the noise field below 5 Hz is controlled by nonlinear wave–wave interactions and that existing theories account adequately for the effects observed.

cant papers; those commented upon below happen to be more closely related to my own current interests, and the choice does not imply that the others are not worth equal attention.In a chapter entitled "Review of Selected Models of Speech Perception" (Chap. 6), Dennis Klatt presents a masterly overview of selected models of lexical hypothesis formation from acoustic data. Klatt offers both a description and a reasoned critique of the following models: a revised version of the motor theory (Liberman and Mattingly, 1986); analysis by synthesis using broad phonetic classification for initial lexical search (Zue, 1986); lexical access from phonetic features (Stevens, 1986); the LAFS (Lexical Access from Spectra) model of spectral sequences for words (Klatt 1979(Klatt , 1986; auditory front ends that attempt to model the encoding of speech by the peripheral auditory system; pseudoneural models and other network models such as Trace (Elman and McClelland, 1986); phonetic recognition using features as input to a decision strategy involving fuzzy logical prototypes (Massaro, 1987); the perceptual-pointer "speech is not special" theory (Miller, 1982); and hidden Markov models (Jelinek, 1985).Not surprisingly, Klatt finds his own LAPS model most attractive, since "all the detailed knowledge about the relations between acoustics and word sequences is contained in a network of expected spectral sequences" (p. 217). He grants, however, that it has not been proved that enumeration of alternatives is a feasible approach to the characterization of variability inherent in speech. Klatt recognizes that LAPS is basically an atheoretical approach, and admits that until such time as a working system is constructed, LAPS is no more convincing than several other candidate models.In general, theories of lexical access seem to assume that word boundaries are somehow "given," as they are marked by spaces in a printed text. Klatt's model makes provisions for word boundary detection by assuming, first, that matching at the lexical level is done in terms of spectra, that any phonetic transition can be characterized by a sequence of spectra (or by several alternative spectral sequences), and that phonetic recoding across word boundaries is best represented by a decoding network that enumerates all possible spectral sequences for all possible word combinations (of English) (pp. 193-194). Klatt's model, and most of the others reviewed in the article, are basi-J.