A Population Rate Code of Auditory Space in the Human Cortex

Background

Previous work on the human auditory cortex has revealed areas specialized in spatial processing but how the neurons in these areas represent the location of a sound source remains unknown.

Methodology/Principal Findings

Here, we performed a magnetoencephalography (MEG) experiment with the aim of revealing the neural code of auditory space implemented by the human cortex. In a stimulus-specific adaptation paradigm, realistic spatial sound stimuli were presented in pairs of adaptor and probe locations. We found that the attenuation of the N1m response depended strongly on the spatial arrangement of the two sound sources. These location-specific effects showed that sounds originating from locations within the same hemifield activated the same neuronal population regardless of the spatial separation between the sound sources. In contrast, sounds originating from opposite hemifields activated separate groups of neurons.

Conclusions/Significance

These results are highly consistent with a rate code of spatial location formed by two opponent populations, one tuned to locations in the left and the other to those in the right. This indicates that the neuronal code of sound source location implemented by the human auditory cortex is similar to that previously found in other primates.     

Discrimination with a Mixture Population

The regions of classification when one of the populations is a mixture of two or more multivariate normal populations are found to be defined by the functions of the linear discriminant functions used for discriminating between the components of the mixture population and the other multivariate normal population.     

Effects of Presentation Rate and Attention on Auditory Discrimination: A Comparison of Long-Latency Auditory Evoked Potentials in School-Aged Children and Adults

Decoding human speech requires both perception and integration of brief, successive auditory stimuli that enter the central nervous system as well as the allocation of attention to language-relevant signals. This study assesses the role of attention on processing rapid transient stimuli in adults and children. Cortical responses (EEG/ERPs), specifically mismatch negativity (MMN) responses, to paired tones (standard 100–100Hz; deviant 100–300Hz) separated by a 300, 70 or 10ms silent gap (ISI) were recorded under Ignore and Attend conditions in 21 adults and 23 children (6–11 years old). In adults, an attention-related enhancement was found for all rate conditions and laterality effects (L>R) were observed. In children, 2 auditory discrimination-related peaks were identified from the difference wave (deviant-standard): an early peak (eMMN) at about 100–300ms indexing sensory processing, and a later peak (LDN), at about 400–600ms, thought to reflect reorientation to the deviant stimuli or “second-look” processing. Results revealed differing patterns of activation and attention modulation for the eMMN in children as compared to the MMN in adults: The eMMN had a more frontal topography as compared to adults and attention played a significantly greater role in childrens’ rate processing. The pattern of findings for the LDN was consistent with hypothesized mechanisms related to further processing of complex stimuli. The differences between eMMN and LDN observed here support the premise that separate cognitive processes and mechanisms underlie these ERP peaks. These findings are the first to show that the eMMN and LDN differ under different temporal and attentional conditions, and that a more complete understanding of children’s responses to rapid successive auditory stimulation requires an examination of both peaks.     

Extrapolation from Individual-Level Responses to Population Growth Rate Using Population Modeling

Ecological risk assessments of chemicals are often based on simple measurements of toxicity in individuals. However, the protection goals are often set at the population and community levels. Population models may be a useful tool to extrapolate from individual-level measurements to population-level endpoints. In the present study, the population growth rate (λ) was calculated for three sets of full life-cycle data (Tetranychus urticae exposed to agrimek, and Daphnia pulex exposed to spinosad and diazinon). The results were compared to λ from population models, where survival and/or reproduction were adjusted according to 4 d of data from the same life-cycle data. This was done to determine whether truncated demographic data can give results similar to that obtained with full life-cycle data. The resulting correlations were strong when both effects on survival and reproduction were included in the model (p < .001, 0.93 < R₂ < 1.00). There were also strong correlations in several cases when only effects on survival or reproduction were considered, although the total risk to the population tended to be underestimated. The results of the present study show that population models can be useful to extrapolate truncated data on the individual level to more ecologically relevant population-level endpoints.     

The Neural Code for Auditory Space Depends on Sound Frequency and Head Size in an Optimal Manner

A major cue to the location of a sound source is the interaural time difference (ITD)–the difference in sound arrival time at the two ears. The neural representation of this auditory cue is unresolved. The classic model of ITD coding, dominant for a half-century, posits that the distribution of best ITDs (the ITD evoking a neuron’s maximal response) is unimodal and largely within the range of ITDs permitted by head-size. This is often interpreted as a place code for source location. An alternative model, based on neurophysiology in small mammals, posits a bimodal distribution of best ITDs with exquisite sensitivity to ITDs generated by means of relative firing rates between the distributions. Recently, an optimal-coding model was proposed, unifying the disparate features of these two models under the framework of efficient coding by neural populations. The optimal-coding model predicts that distributions of best ITDs depend on head size and sound frequency: for high frequencies and large heads it resembles the classic model, for low frequencies and small head sizes it resembles the bimodal model. The optimal-coding model makes key, yet unobserved, predictions: for many species, including humans, both forms of neural representation are employed, depending on sound frequency. Furthermore, novel representations are predicted for intermediate frequencies. Here, we examine these predictions in neurophysiological data from five mammalian species: macaque, guinea pig, cat, gerbil and kangaroo rat. We present the first evidence supporting these untested predictions, and demonstrate that different representations appear to be employed at different sound frequencies in the same species.     

Mechanisms of Selective Attention during Competitive Discrimination of Visual and Auditory Verbal Information: Positron Emission Tomography and Cortical Evoked Potential Studies

The mechanisms of selective verbal attention were studied under conditions of simultaneous delivery of speech signals via the visual and auditory channels. The investigation was based on the comparison and synthesis of data obtained by two methods: positron emission tomography (PET) and brain evoked potentials (EPs). A new approach was developed: complementary tasks were constructed in such a way that, despite principal methodological problems, the same phenomenon could be investigated in one paradigm in EP and PET studies. The results obtained by the two methods are in rather good agreement with respect to topography: the secondary and tertiary areas, as well as the associative brain areas, are involved in attention concentration, that is, selection of verbal information occurs at the level of cognitive processes. The combination of two complementary methods, PET and EP, allowed the processes of processing of sensory information and brain mechanisms of selective attention to be investigated much more completely. The PET studies contributed to further understanding of brain mechanisms evidencing where processing occurs and the EP method provided insight into the mechanism of how this information is processed inside the corresponding cortical areas. The finding that the activation of primary areas of the visual cortex is accompanied by the inhibition of visual information deserves attention. This conclusion can be considered highly significant because of the concordance of the two independent methods. How to interpret it is not yet clear. It is possible that, in the case of primary importance of verbal information and priority of the visual channel for the repression from consciousness of artificially irrelevant information, a safety mechanism is activated: the amplified signal enters the brain cortex, where it is retained in the short-term iconic memory. This enables a reaction to this stimulus (if necessary), in the presence of any additional sign involving selective attention.     

Mapping Cortical Dynamics Using Simultaneous MEG/EEG and Anatomically-constrained Minimum-norm Estimates: an Auditory Attention Example

Magneto- and electroencephalography (MEG/EEG) are neuroimaging techniques that provide a high temporal resolution particularly suitable to investigate the cortical networks involved in dynamical perceptual and cognitive tasks, such as attending to different sounds in a cocktail party. Many past studies have employed data recorded at the sensor level only, i.e., the magnetic fields or the electric potentials recorded outside and on the scalp, and have usually focused on activity that is time-locked to the stimulus presentation. This type of event-related field / potential analysis is particularly useful when there are only a small number of distinct dipolar patterns that can be isolated and identified in space and time. Alternatively, by utilizing anatomical information, these distinct field patterns can be localized as current sources on the cortex. However, for a more sustained response that may not be time-locked to a specific stimulus (e.g., in preparation for listening to one of the two simultaneously presented spoken digits based on the cued auditory feature) or may be distributed across multiple spatial locations unknown a priori, the recruitment of a distributed cortical network may not be adequately captured by using a limited number of focal sources.Here, we describe a procedure that employs individual anatomical MRI data to establish a relationship between the sensor information and the dipole activation on the cortex through the use of minimum-norm estimates (MNE). This inverse imaging approach provides us a tool for distributed source analysis. For illustrative purposes, we will describe all procedures using FreeSurfer and MNE software, both freely available. We will summarize the MRI sequences and analysis steps required to produce a forward model that enables us to relate the expected field pattern caused by the dipoles distributed on the cortex onto the M/EEG sensors. Next, we will step through the necessary processes that facilitate us in denoising the sensor data from environmental and physiological contaminants. We will then outline the procedure for combining and mapping MEG/EEG sensor data onto the cortical space, thereby producing a family of time-series of cortical dipole activation on the brain surface (or "brain movies") related to each experimental condition. Finally, we will highlight a few statistical techniques that enable us to make scientific inference across a subject population (i.e., perform group-level analysis) based on a common cortical coordinate space.     

Genetic Pleiotropy Explains Associations between Musical Auditory Discrimination and Intelligence

Musical aptitude is commonly measured using tasks that involve discrimination of different types of musical auditory stimuli. Performance on such different discrimination tasks correlates positively with each other and with intelligence. However, no study to date has explored these associations using a genetically informative sample to estimate underlying genetic and environmental influences. In the present study, a large sample of Swedish twins (N = 10,500) was used to investigate the genetic architecture of the associations between intelligence and performance on three musical auditory discrimination tasks (rhythm, melody and pitch). Phenotypic correlations between the tasks ranged between 0.23 and 0.42 (Pearson r values). Genetic modelling showed that the covariation between the variables could be explained by shared genetic influences. Neither shared, nor non-shared environment had a significant effect on the associations. Good fit was obtained with a two-factor model where one underlying shared genetic factor explained all the covariation between the musical discrimination tasks and IQ, and a second genetic factor explained variance exclusively shared among the discrimination tasks. The results suggest that positive correlations among musical aptitudes result from both genes with broad effects on cognition, and genes with potentially more specific influences on auditory functions.     

Neuronal Adaptation Translates Stimulus Gaps into a Population Code

Neurons in sensory pathways exhibit a vast multitude of adaptation behaviors, which are assumed to aid the encoding of temporal stimulus features and provide the basis for a population code in higher brain areas. Here we study the transition to a population code for auditory gap stimuli both in neurophysiological recordings and in a computational network model. Independent component analysis (ICA) of experimental data from the inferior colliculus of Mongolian gerbils reveals that the network encodes different gap sizes primarily with its population firing rate within 30 ms after the presentation of the gap, where longer gap size evokes higher network activity. We then developed a computational model to investigate possible mechanisms of how to generate the population code for gaps. Phenomenological (ICA) and functional (discrimination performance) analyses of our simulated networks show that the experimentally observed patterns may result from heterogeneous adaptation, where adaptation provides gap detection at the single neuron level and neuronal heterogeneity ensures discriminable population codes for the whole range of gap sizes in the input. Furthermore, our work suggests that network recurrence additionally enhances the network''s ability to provide discriminable population patterns.     

Auditory Mismatch Negativity and Repetition Suppression Deficits in Schizophrenia Explained by Irregular Computation of Prediction Error

BackgroundThe predictive coding model is rapidly gaining attention in schizophrenia research. It posits the neuronal computation of residual variance (‘prediction error’) between sensory information and top-down expectation through multiple hierarchical levels. Event-related potentials (ERP) reflect cortical processing stages that are increasingly interpreted in the light of the predictive coding hypothesis. Both mismatch negativity (MMN) and repetition suppression (RS) measures are considered a prediction error correlates based on error detection and error minimization, respectively.MethodsTwenty-five schizophrenia patients and 25 healthy controls completed auditory tasks designed to elicit MMN and RS responses that were investigated using repeated measures models and strong spatio-temporal a priori hypothesis based on previous research. Separate correlations were performed for controls and schizophrenia patients, using age and clinical variables as covariates.ResultsMMN and RS deficits were largely replicated in our sample of schizophrenia patients. Moreover, MMN and RS measures were strongly correlated in healthy controls, while no correlation was found in schizophrenia patients. Single-trial analyses indicated significantly lower signal-to-noise ratio during prediction error computation in schizophrenia.ConclusionsThis study provides evidence that auditory ERP components relevant for schizophrenia research can be reconciled in the light of the predictive coding framework. The lack of any correlation between the investigated measures in schizophrenia patients suggests a disruption of predictive coding mechanisms in general. More specifically, these results suggest that schizophrenia is associated with an irregular computation of residual variance between sensory input and top-down models, i.e. prediction error.     

Auditory Spatial Discrimination and the Mismatch Negativity Response in Hearing-Impaired Individuals

The aims of the present study were to investigate the ability of hearing-impaired (HI) individuals with different binaural hearing conditions to discriminate spatial auditory-sources at the midline and lateral positions, and to explore the possible central processing mechanisms by measuring the minimal audible angle (MAA) and mismatch negativity (MMN) response. To measure MAA at the left/right 0°, 45° and 90° positions, 12 normal-hearing (NH) participants and 36 patients with sensorineural hearing loss, which included 12 patients with symmetrical hearing loss (SHL) and 24 patients with asymmetrical hearing loss (AHL) [12 with unilateral hearing loss on the left (UHLL) and 12 with unilateral hearing loss on the right (UHLR)] were recruited. In addition, 128-electrode electroencephalography was used to record the MMN response in a separate group of 60 patients (20 UHLL, 20 UHLR and 20 SHL patients) and 20 NH participants. The results showed MAA thresholds of the NH participants to be significantly lower than the HI participants. Also, a significantly smaller MAA threshold was obtained at the midline position than at the lateral position in both NH and SHL groups. However, in the AHL group, MAA threshold for the 90° position on the affected side was significantly smaller than the MMA thresholds obtained at other positions. Significantly reduced amplitudes and prolonged latencies of the MMN were found in the HI groups compared to the NH group. In addition, contralateral activation was found in the UHL group for sounds emanating from the 90° position on the affected side and in the NH group. These findings suggest that the abilities of spatial discrimination at the midline and lateral positions vary significantly in different hearing conditions. A reduced MMN amplitude and prolonged latency together with bilaterally symmetrical cortical activations over the auditory hemispheres indicate possible cortical compensatory changes associated with poor behavioral spatial discrimination in individuals with HI.     

Selective Impairment in Frequency Discrimination in a Mouse Model of Tinnitus

Tinnitus is an auditory disorder, which affects millions of Americans, including active duty service members and veterans. It is manifested by a phantom sound that is commonly restricted to a specific frequency range. Because tinnitus is associated with hearing deficits, understanding how tinnitus affects hearing perception is important for guiding therapies to improve the quality of life in this vast group of patients. In a rodent model of tinnitus, prolonged exposure to a tone leads to a selective decrease in gap detection in specific frequency bands. However, whether and how hearing acuity is affected for sounds within and outside those frequency bands is not well understood. We induced tinnitus in mice by prolonged exposure to a loud mid-range tone, and behaviorally assayed whether mice exhibited a change in frequency discrimination acuity for tones embedded within the mid-frequency range and high-frequency range at 1, 4, and 8 weeks post-exposure. A subset of tone-exposed mice exhibited tinnitus-like symptoms, as demonstrated by selective deficits in gap detection, which were restricted to the high frequency range. These mice exhibited impaired frequency discrimination both for tones in the mid-frequency range and high-frequency range. The remaining tone exposed mice, which did not demonstrate behavioral evidence of tinnitus, showed temporary deficits in frequency discrimination for tones in the mid-frequency range, while control mice remained unimpaired. Our findings reveal that the high frequency-specific deficits in gap detection, indicative of tinnitus, are associated with impairments in frequency discrimination at the frequency of the presumed tinnitus.     

Auditory Processing of Temporal Information in a Neotropical Frog is Independent of Signal Intensity

Animals often encounter the problem of identifying the temporal structure of their species-specific communication sounds amidst heterospecific signals that are more intense. Little information is available, however, concerning the effects of intensity on these discriminative capacities. Here we report that male frogs, Physalaemus pustulosus, are able to discriminate between sounds that differ only in their direction of frequency modulation and that this discrimination remains intact over their entire response range; more than a 10,000-fold range in sound intensity in some cases.     

Common Cortical Loci Are Activated during Visuospatial Interpolation and Orientation Discrimination Judgements

There is a wealth of literature on the role of short-range interactions between low-level orientation-tuned filters in the perception of discontinuous contours. However, little is known about how spatial information is integrated across more distant regions of the visual field in the absence of explicit local orientation cues, a process referred to here as visuospatial interpolation (VSI). To examine the neural correlates of VSI high field functional magnetic resonance imaging was used to study brain activity while observers either judged the alignment of three Gabor patches by a process of interpolation or discriminated the local orientation of the individual patches. Relative to a fixation baseline the two tasks activated a largely over-lapping network of regions within the occipito-temporal, occipito-parietal and frontal cortices. Activated clusters specific to the orientation task (orientation>interpolation) included the caudal intraparietal sulcus, an area whose role in orientation encoding per se has been hotly disputed. Surprisingly, there were few task-specific activations associated with visuospatial interpolation (VSI>orientation) suggesting that largely common cortical loci were activated by the two experimental tasks. These data are consistent with previous studies that suggest higher level grouping processes -putatively involved in VSI- are automatically engaged when the spatial properties of a stimulus (e.g. size, orientation or relative position) are used to make a judgement.     

A Common Cortical Circuit Mechanism for Perceptual Categorical Discrimination and Veridical Judgment

Perception involves two types of decisions about the sensory world: identification of stimulus features as analog quantities, or discrimination of the same stimulus features among a set of discrete alternatives. Veridical judgment and categorical discrimination have traditionally been conceptualized as two distinct computational problems. Here, we found that these two types of decision making can be subserved by a shared cortical circuit mechanism. We used a continuous recurrent network model to simulate two monkey experiments in which subjects were required to make either a two-alternative forced choice or a veridical judgment about the direction of random-dot motion. The model network is endowed with a continuum of bell-shaped population activity patterns, each representing a possible motion direction. Slow recurrent excitation underlies accumulation of sensory evidence, and its interplay with strong recurrent inhibition leads to decision behaviors. The model reproduced the monkey's performance as well as single-neuron activity in the categorical discrimination task. Furthermore, we examined how direction identification is determined by a combination of sensory stimulation and microstimulation. Using a population-vector measure, we found that direction judgments instantiate winner-take-all (with the population vector coinciding with either the coherent motion direction or the electrically elicited motion direction) when two stimuli are far apart, or vector averaging (with the population vector falling between the two directions) when two stimuli are close to each other. Interestingly, for a broad range of intermediate angular distances between the two stimuli, the network displays a mixed strategy in the sense that direction estimates are stochastically produced by winner-take-all on some trials and by vector averaging on the other trials, a model prediction that is experimentally testable. This work thus lends support to a common neurodynamic framework for both veridical judgment and categorical discrimination in perceptual decision making.     

Multiplicative Auditory Spatial Receptive Fields Created by a Hierarchy of Population Codes

A multiplicative combination of tuning to interaural time difference (ITD) and interaural level difference (ILD) contributes to the generation of spatially selective auditory neurons in the owl''s midbrain. Previous analyses of multiplicative responses in the owl have not taken into consideration the frequency-dependence of ITD and ILD cues that occur under natural listening conditions. Here, we present a model for the responses of ITD- and ILD-sensitive neurons in the barn owl''s inferior colliculus which satisfies constraints raised by experimental data on frequency convergence, multiplicative interaction of ITD and ILD, and response properties of afferent neurons. We propose that multiplication between ITD- and ILD-dependent signals occurs only within frequency channels and that frequency integration occurs using a linear-threshold mechanism. The model reproduces the experimentally observed nonlinear responses to ITD and ILD in the inferior colliculus, with greater accuracy than previous models. We show that linear-threshold frequency integration allows the system to represent multiple sound sources with natural sound localization cues, whereas multiplicative frequency integration does not. Nonlinear responses in the owl''s inferior colliculus can thus be generated using a combination of cellular and network mechanisms, showing that multiple elements of previous theories can be combined in a single system.     

Different Stimuli,Different Spatial Codes: A Visual Map and an Auditory Rate Code for Oculomotor Space in the Primate Superior Colliculus

Maps are a mainstay of visual, somatosensory, and motor coding in many species. However, auditory maps of space have not been reported in the primate brain. Instead, recent studies have suggested that sound location may be encoded via broadly responsive neurons whose firing rates vary roughly proportionately with sound azimuth. Within frontal space, maps and such rate codes involve different response patterns at the level of individual neurons. Maps consist of neurons exhibiting circumscribed receptive fields, whereas rate codes involve open-ended response patterns that peak in the periphery. This coding format discrepancy therefore poses a potential problem for brain regions responsible for representing both visual and auditory information. Here, we investigated the coding of auditory space in the primate superior colliculus(SC), a structure known to contain visual and oculomotor maps for guiding saccades. We report that, for visual stimuli, neurons showed circumscribed receptive fields consistent with a map, but for auditory stimuli, they had open-ended response patterns consistent with a rate or level-of-activity code for location. The discrepant response patterns were not segregated into different neural populations but occurred in the same neurons. We show that a read-out algorithm in which the site and level of SC activity both contribute to the computation of stimulus location is successful at evaluating the discrepant visual and auditory codes, and can account for subtle but systematic differences in the accuracy of auditory compared to visual saccades. This suggests that a given population of neurons can use different codes to support appropriate multimodal behavior.     

Estimating Selection Intensity on Synonymous Codon Usage in a Nonequilibrium Population

Codon usage bias is the nonrandom use of synonymous codons for the same amino acid. Most population genetic models of codon usage evolution assume that the population is at mutation–selection–drift equilibrium. Natural populations, however, frequently deviate from equilibrium, often because of recent demographic changes. Here, we construct a matrix model that includes the effects of a recent change in population size on estimates of selection on preferred vs. unpreferred codons. Our results suggest that patterns of synonymous polymorphisms affecting codon usage can be quite erratic after such a change; statistical methods that fail to take demographic effects into account can then give incorrect estimates of important parameters. We propose a new method that can accurately estimate both demographic and codon usage parameters. The method also provides a simple way of testing for the effects of covariates such as gene length and level of gene expression on the intensity of selection, which we apply to a large Drosophila melanogaster polymorphism data set. Our analyses of twofold degenerate codons reveal that (i) selection acts in favor of preferred codons, (ii) there is mutational bias in favor of unpreferred codons, (iii) shorter genes and genes with higher expression levels are under stronger selection, and (iv) there is little evidence for a recent change in population size in the Zimbabwe population of D. melanogaster.CODONS specifying the same amino acid are called synonymous codons. These are often used nonrandomly, with some codons appearing more frequently than others. This biased usage of synonymous codons has been found in many organisms such as Drosophila, yeast, and bacteria (Ikemura 1985; Duret and Mouchiroud 1999; Hershberg and Petrov 2008). Conventionally, synonymous codons for a given amino acid are divided into two classes: preferred and unpreferred codons (Ikemura 1985; Akashi 1994; Duret and Mouchiroud 1999). Several observations indicate that codon usage is affected by natural selection. First, in species with codon usage bias, preferred codons generally correspond to the most abundant tRNA species (Ikemura 1981). Second, highly expressed genes usually have higher codon usage bias than genes with low expression (Sharp and Li 1986; Duret and Mouchiroud 1999; Hey and Kliman 2002). Third, the synonymous substitution rate of a gene has been shown to be negatively correlated with its degree of codon usage bias (Sharp and Li 1986; Bierne and Eyre-Walker 2006). The most commonly cited explanations of the apparent fitness differences between preferred and unpreferred codons are selection for translation efficiency, translational accuracy, and mRNA stability (Ikemura 1985; Eyre-Walker and Bulmer 1993; Akashi 1994; Drummond et al. 2005). Recently, it has been proposed that exon splicing also affects codon usage bias (Warnecke and Hurst 2007).From a population genetics perspective, the extent of codon usage bias is ultimately a product of the joint effects of mutation, selection, genetic drift, recombination, and demographic history. The Li–Bulmer model of drift, selection, and reversible mutation between preferred and unpreferred codons at a site is the most widely used model (Li 1987; Bulmer 1991; McVean and Charlesworth 1999). Applications of this model generally assume that the population is at mutation–selection–drift equilibrium. However, empirical studies have suggested that changes in the strengths of various driving forces may not be unusual. For example, in Drosophila melanogaster, there is evidence that the population size (Li and Stephan 2006; Thornton and Andolfatto 2006; Keightley and Eyre-Walker 2007; Stephan and Li 2007), recombinational landscape (Takano-Shimizu 1999), and mutational process (Takano-Shimizu 2001; Kern and Begun 2005) may have changed significantly over the species'' evolutionary history.Such changes cause departures from equilibrium. Theoretical models show that it takes a very long time, proportional to the reciprocal of the mutation rate, for the population to approach a new equilibrium state (Tachida 2000; Comeron and Kreitman 2002). Before reaching equilibrium, the population often shows counterintuitive patterns of evolution (Eyre-Walker 1997; Takano-Shimizu 1999, 2001; Comeron and Kreitman 2002; Comeron and Guthrie 2005; Charlesworth and Eyre-Walker 2007). Despite these theoretical results, details of the patterns of polymorphism and substitution rates following a recent change in population size, and their effects on estimates of strength of selection, have not been determined.The above findings point to the importance of incorporating nonequilibrium factors into the study of codon usage bias. To this end, we extend the Li–Bulmer model to allow population size to vary over time, by representing the evolutionary process by a transition matrix. By analyzing this matrix model, we show that a recent change in population size can result in erratic patterns of codon usage and that methods failing to take into account these demographic effects can give false estimates of the intensity of selection.To solve these problems, we propose a new method, which does not require polarizing ancestral vs. derived states using outgroup data (cf. Cutter and Charlesworth 2006), but requires only knowledge of preferred vs. unpreferred states defined by patterns of codon usage. We use information on both polymorphic and fixed sites, which enables both mutational bias and the strength of selection to be estimated, in contrast to previous methods that use information on polymorphisms alone. Simulations indicate that this method can accurately estimate both demographic and codon usage parameters and can distinguish between selection and demography. We use the new method to analyze a large D. melanogaster polymorphism data set (Shapiro et al. 2007) and find evidence for natural selection on synonymous codons. We use our approach to show that genes with shorter coding sequences and higher levels of expression are under significantly stronger selection than longer genes with lower expression.