We analyze new optical spectra of a sample of 502 galaxies in close pairs and n-tuples, separated by <= 50/h kpc. We extracted the sample objectively from the CfA2 redshift survey, without regard to the surroundings of the tight systems. We probe the relationship between star formation and the dynamics of the systems of galaxies. The equivalent widths of Hα(EW(Hα) and other emission lines anti-correlate strongly with pair spatial separation (ΔD) and velocity separation. We use the measured EW(Hα) and the starburst models of Leitherer et al. to estimate the time since the most recent burst of star for- mation began for each galaxy. In the absence of a large contribution from an old stellar population to the continuum around Hα, the observed ΔD -- EW(Hα) correlation signifies that starbursts with larger separations on the sky are, on average, older. By matching the dynamical timescale to the burst timescale, we show that the data support a simple picture in which a close pass initiates a starburst; EW(Hα) decreases with time as the pair separation increases, accounting for the anti-correlation. This picture leads to a method for measuring the duration and the initial mass function of interaction-induced starbursts: our data are compatible with the starburst and orbit models in many respects, as long as the starburst lasts longer than \sim10^8 years and the delay between the close pass and the initiation of the starburst is less than a few \times 10^7 years. If there is no large contribution from an old stellar population to the continuum around Hαthe Miller-Scalo and cutoff (M <= 30 M_\sun) Salpeter initial mass functions fit the data much better than a standard Salpeter IMF. (Abridged.)

43 pages, 22 figures, to appear in the ApJ; we correct an error which had minor effects on numerical values in the paper