Changes in the neural control of a complex motor sequence during learning

Abstract

The acquisition of complex motor sequences often proceeds through trial-and-error learning, requiring the deliberate exploration of motor actions and the concomitant evaluation of the resulting performance. Songbirds learn their song in this manner, producing highly variable vocalizations as juveniles. As the song improves, vocal variability is gradually reduced until it is all but eliminated in adult birds. In the present study we examine how the motor program underlying such a complex motor behavior evolves during learning by recording from the robust nucleus of the arcopallium (RA), a motor cortex analog brain region. In young birds, neurons in RA exhibited highly variable firing patterns that throughout development became more precise, sparse, and bursty. We further explored how the developing motor program in RA is shaped by its two main inputs: LMAN, the output nucleus of a basal ganglia-forebrain circuit, and HVC, a premotor nucleus. Pharmacological inactivation of LMAN during singing made the song-aligned firing patterns of RA neurons adultlike in their stereotypy without dramatically affecting the spike statistics or the overall firing patterns. Removing the input from HVC, on the other hand, resulted in a complete loss of stereotypy of both the song and the underlying motor program. Thus our results show that a basal ganglia-forebrain circuit drives motor exploration required for trial-and-error learning by adding variability to the developing motor program. As learning proceeds and the motor circuits mature, the relative contribution of LMAN is reduced, allowing the premotor input from HVC to drive an increasingly stereotyped song.

Fig. 1.

Development of the robust nucleus of the arcopallium (RA) motor program in zebra finches. A: diagram showing the main pathways involved in song learning and song production. The motor pathway (gray) includes motor cortex analogs HVC and RA, whereas the anterior forebrain pathway (AFP; white), a basal ganglia thalamocortical circuit, consists of area X, the dorsolateral anterior thalamic nucleus (DLM), and the lateral magnocellular nucleus of the anterior nidopallium (LMAN), which, in turn, projects to RA. Chronic single-unit recordings in RA were carried out throughout sensorimotor learning and in adult birds. Syrinx, vocal organ; nXIIts, tracheosyringeal portion of the hypoglossal nucleus. B: development of song and the underlying RA firing patterns throughout song learning. Song spectrograms and corresponding raw traces of recorded RA neurons are shown at different developmental stages, subsong, plastic song, and adult song, in 3 different birds.

Fig. 2.

Song-aligned firing patterns in RA neurons change gradually during learning. As the song becomes less variable and more similar to the tutor, the firing patterns in RA become more reproducible, sparse, and bursty. Raster plots show 5 representative RA projection neurons recorded at 5 different developmental time points. Each color represents a cell, and each row represents a rendition of the motif. Dark bars above the spike rasters indicate the beginning and end of identified syllables within each motif. For each developmental time point, the cells were recorded serially (i.e., not simultaneously). dph, Days posthatch.

Fig. 3.

Song learning, characterized by a decrease in song variability, is accompanied by changes in the statistics of RA projection neuron firing. A: acoustic variability as a function of age. Mean acoustic variability scores of readily identifiable syllables were calculated for 4 birds and averaged across birds. Shaded area denotes confidence interval (P < 0.05). B: interspike interval (ISI) distribution for birds at different ages throughout song development, corresponding roughly to subsong/early plastic song (43–49 dph), mid-plastic song (58–64 dph), late plastic song (79–87 dph), young adult song (115–160 dph), and older adult directed song (200+ dph). All ISI distributions were smoothed with an 8-ms square window. C: the fraction of spikes in bursts (>150 Hz). D: average firing rate during singing. E: sparseness index (see methods). F: average pairwise correlation of song-aligned spike trains. Black circles represent cells; red squares represent the average for different age groups described in B. (For E and F, the first point represents early plastic song at ages 50–51 dph; see methods).

Fig. 4.

Transient inactivation of LMAN decreases variability of RA activity patterns and vocal output. A: experimental setup. Recordings were made from RA neurons while LMAN was inactivated using reverse microdialysis (see methods). B: behavioral effect on acoustic variability following infusion of GABA and lidocaine into the dialysis probes implanted bilaterally in LMAN. Each data point was calculated from 10 renditions of 4 identified syllables (see methods) in a single bird (73 and 74 dph for GABA and lidocaine, respectively). C: 8 consecutive renditions of a syllable before (LMAN intact) and during 2% lidocaine infusion (LMAN inactivated). D–F, top: song-aligned raster plots of spike trains of 2 representative RA projection neurons (D and E) and 1 interneuron (F). Time advances from top to bottom; red bar indicates recordings during LMAN inactivation. To better contrast the 2 conditions, the period of drug wash-in (0–20 min after start of infusion) is not shown. Other than for the period of drug infusion, the shown spike trains were recorded during consecutive renditions of the song motif. Asterisks denote the start and end of the song motif to which the spike trains were aligned. Below the spike trains are the song-aligned histograms for the neurons. D–F, bottom: changes in the ISI distribution as a consequence of LMAN inactivation (red line, LMAN inactivated; blue line, LMAN intact). As is evident from the spike rasters above, LMAN can introduce high-frequency bursts into the RA firing patterns, and thus LMAN inactivation shifts the distribution toward longer ISIs. Reduction in high-frequency bursts (>400 Hz) for the 3 cells shown was 43, 21, and 90%, respectively. TTX, tetrodotoxin.

Fig. 5.

Effects of LMAN inactivation on the RA motor program and song. A: matrix of pairwise spike train cross-correlations across different renditions for the neuron represented in Fig. 4D. Note the increased correlation after LMAN inactivation (red bar). B: matrix of pairwise song similarity scores across different renditions, corresponding to the motifs in A. The bird sang throughout the drug wash-in period (denoted by arrow), and the gradual shift toward more stereotypy can be seen in both the acoustic similarity and spike correlation. C: average spike train correlation for the RA cell population with and without LMAN active. Referencing the matrix in A, the graph depicts the average correlation in quadrant 1 vs. quadrant 2 for all neurons. D: pairwise correlation between spike trains in RA with LMAN intact and spike trains with LMAN silenced (quadrant 3 in A) is higher than the correlation between spike trains with LMAN intact (quadrant 1 in A). E: the difference in ISI distribution across the population of projection neurons (n = 18 cells) with and without LMAN active reveals a net increase in short ISIs with LMAN activity. Shaded area denotes confidence interval (P < 0.05). ISI distributions were smoothed with an 8-ms square window. F: average firing rate for projection neurons is similar with and without LMAN active. For C, D, and F, black data points denote the projection neurons in Fig 4, the whereas the green data point represents the interneuron. Error bars denote SD.

Fig. 6.

Firing patterns in RA projection neurons revert to a subsonglike state following the elimination of inputs from HVC. A: recording of a single neuron in RA (black trace) during undirected singing in an adult bird. B: recording of an RA neuron during undirected singing in an adult bird following complete bilateral transection of the HVC-to-RA fiber tract. C: ISI distributions of neurons recorded in intact adult birds, subsong-producing birds, and adult birds following transection. D: mean firing rate of each neuron recorded in intact adult birds, subsong birds, and adult bird following transection. E: fraction of spikes contained in bursts (>150 Hz instantaneous firing rate). F: duration of spiking suppression after singing. Each data point is the median time of the first spike after bout offset.