The characteristics of the emission-line regions and the derived spectrophotometric parameters are given in Table 6. The galaxy (Mrk) number is given in Col. 1. The label and code for the position of the emission-line regions (1 = nuclear, 0 = extranuclear) are repeated in Cols. 2 and 3. The distance of the galaxy, estimated using a Hubble constant of 75 km s^-1Mpc^-1, is given in Col. 4. The distance of the extranuclear regions from the nucleus, and the diameter of all regions, expressed in kpc, are given in Cols. 5 and 6. The reddening coefficient $c_{\rm H\beta}$ is given in Col. 7. The electronic density ( $n_{\rm e}$ ) is given in Col. 8. The line-intensity ratios, [NII] $\lambda$ 6583/H $\alpha$ , [SII] $\lambda\lambda$ 6716,6731/H $\alpha$ and [OIII] $\lambda$ 5007/H $\beta$ , used for the spectral classification and corrected both for internal reddening and for Balmer absorption, are reported in Cols. 9, 10 and 11, followed by the results of the spectral classification in Col. 12. The H $\alpha$ flux [ $I(\rm{H\alpha})$ in 10^-14 erg s^-1 cm^-2] is given in Col. 13 and the H $\alpha$ luminosity [ $L(\rm{H\alpha}$ ) in erg s^-1] in Col. 14. In the subsequent analysis we only take into account the H $\alpha$ luminosities corrected for reddening, that is, those of galaxies for which we derived the reddening parameter $c_{\rm H\beta}$ (see Sect. 5.3). For the others, we give lower limits on the H $\alpha$ flux and luminosity in Table 6, which is given in electronic form only.

At this stage, it is necessary to clarify the definition of the different types of starburst galaxies which will be mentioned in this paper. One can first separate HII galaxies and starburst galaxies by the properties of the host galaxy. HII galaxies are mainly metal-poor dwarf irregular or blue compact galaxies with low dust content (e.g. French 1980; Keel 1983 and Terlevich et al.1991). They contain many giant HII regions with high-excitation spectra, with properties very close to those of extragalactic giant HII regions distributed in the disk of nearby spiral galaxies (e.g. McCall et al. 1985). HII nucleus galaxies (Stauffer 1982; Keel 1983; Kennicutt et al. 1989; Ho et al. 1997a) and Starburst Nucleus Galaxies (SBNGs; Balzano 1983; Coziol et al.1994) are defined as galaxies with HII regions in their nuclei. They are more massive and chemically evolved spiral galaxies, with a large population of old and evolved stars and a huge quantity of dust (Coziol 1996). The low-excitation spectra of SBNGs reflect the higher metallicity observed in spiral galaxies, and especially in their nucleus, when compared to irregular or compact HII galaxies. A last distinction is made between HII nuclei and SBNGs on the basis of their nuclear H $\alpha$ luminosity, HII nuclei being fainter (L(H $\alpha$ ) $$\mathrel{\mathchoice {\vcenter{\offinterlineskip\halign{\hfil $\displaystyle ...$ erg s^-1) than SBNGs. In terms of star formation, HII nuclei represent the low-luminosity end of SBNGs (Coziol 1996).
Table 6: Derived spectrophotometric parameters (available in electronic form only)

The results of our spectrophotometric analysis are displayed in Fig. 1 which shows the location of the emission-line regions in two diagnostic diagrams of Veilleux & Osterbrock (1987). In each diagram, the continuous curve, empirically derived by Veilleux & Osterbrock (1987), separates starburst nuclei and HII regions where the gas is assumed to be ionized by young stars, from AGNs where the main ionizing source is thought to be an accretion disk around a black hole (e.g. Rees 1984) which produces a power law spectrum. We make a further distinction among AGNs between objects of high ([OIII]/H $\beta$ > 3) and low ([OIII]/H $\beta$ $\leq$ 3) excitation (horizontal line, Shuder & Osterbrock 1981). The first group represents the classical Seyfert 2 galaxies while the LINERs fall in the second group. We did not use the original criteria for defining a LINER (Heckman 1980) because measurements of [OI] $\lambda$ 6300 and [OII] $\lambda$ 3727 were often not available. For the same reason, we did not make use of the [OI] $\lambda$ 6300/H $\alpha$ ratio to distinguish between the different sources of ionization.

**Table 7:** Spectral classification
$\begin{tabular} {llrrr} \hline \hline \multicolumn{2}{c}{Spectral Type$^{\rm a}$... ...dotfill} & 105 (100\%) & 116 (100\%) & 221 (100\%) \\ \hline \hline\end{tabular}$ $^{\rm a}$ SBNG = Starburst Nucleus Galaxy; HIIG = HII galaxy; HII, LINER? = ambiguous classification between HII and LINER (see text for details); Uncertain = classification based on only one emission line ratio ([NII]/H $\alpha$ or [SII]/H $\alpha$ ).

One should note that the classification process is not always unambiguous, for at least two reasons. First, the two conditions involving the low-ionization lines ([NII] and [SII]) do not always hold simultaneously. This reflects the empirical nature of the diagnostic diagrams as well as the possibility that one line ratio is enhanced or depressed with respect to the other one as a result of, for instance, selective abundance variations (see Sect. 5.5.2). Second, large measurement uncertainties may be associated with any given line intensity ratio. Thus, one should evaluate each object individually, taking all of these factors into consideration, before a classification can be assigned to it. When more than one classification is consistent with the data, both are given, with the more likely one listed first (Col. 12 of Table 6). An ambiguous spectral classification (between HII and LINER) arises for 11 nuclear (Mrk 90, 271, 332, 353, 593, 617, 874, 1180, 1200, 1291 and 1485) and 8 extranuclear (Mrk 712-3, 814-1, 814-4, 1086-3, 1302-1, 1363-1, 1363-3 and 1433-3) regions.

The result of the spectral classification is summarized in Table 7. We found that 70 nuclear regions (67% of the sample) have spectra characteristic of photoionization by hot stars, i.e. are classified as starburst nuclei (62%) or HII galaxy (5%). Four galaxies (Mrk 21, 271, 446 and 1452) were classified as SBNGs using only one emission-line ratio ([NII]/H $\alpha$ or [SII]/H $\alpha$ ), their classification is thus rather uncertain. AGN emission lines were observed in 20 nuclei (19%) including 12 Seyfert 1 galaxies (11%) and 8 Seyfert 2 galaxies (8%). Among the 116 extranuclear regions, 81 (69%) are HII regions. Here, the classification is uncertain for 27 regions (24%) and ambiguous between HII and LINER for 8 regions (7%) for the same reason as explained above.

$\begin{figure} \vspace{-0.5cm} \epsfxsize=8.8cm \centering{\mbox{ \epsfbox {ds1465f3.eps} }} \vspace{-3cm}\end{figure}$

Figure 3: Optical spectra of HII galaxies showing strong emission lines and a high excitation ([OIII]/H $\beta$ > 3) compared to starburst nuclei. Intensities are in 10^-14 erg s^-1 cm^-2 Å^-1

$\begin{figure} \vspace{-0.5cm} \epsfxsize=8.8cm \centering{\mbox{ \epsfbox {ds1465f4.eps} }} \vspace{-3cm}\end{figure}$

Figure 4: Optical spectra of Seyfert 1 nuclei. The broad Balmer emission lines (H $\alpha$ , H $\beta$ , etc), compared to the narrower forbidden lines (e.g. [OIII]), allow a clear and rapid classification of these AGNs. Intensities are in 10^-14 erg s^-1 cm^-2 Å^-1

$\begin{figure} \vspace{-0.5cm} \epsfxsize=8.8cm \centering{\mbox{ \epsfbox {ds1465f5.eps} }} \vspace{-3cm}\end{figure}$

Figure 5: Optical spectra of Seyfert 2 nuclei. In these AGNs, the Balmer lines have the same width as the forbidden lines and the excitation is high ([OIII]/H $\beta$ > 3). Intensities are in 10^-14 erg s^-1 cm^-2 Å^-1

$\begin{figure} \vspace{-0.5cm} \epsfxsize=8.8cm \centering{\mbox{ \epsfbox {ds1465f6.eps} }} \vspace{-3cm}\end{figure}$

Figure 6: Examples of objects with an ambiguous classification between HII and LINER. Intensities are in 10^-14 erg s^-1 cm^-2 Å^-1

Representative spectra of each spectral class are shown in Fig. 2 (starburst nuclei), Fig. 3 (HII galaxies), Fig. 4 (Seyfert 1 nuclei), Fig. 5 (Seyfert 2 nuclei), and Fig. 6 (objects with an ambiguous classification).

5.2 Colors

We corrected for reddening the spectral continuum colors (B - V) and (V - R) of the individual regions using the reddening coefficient $c_{\rm H\beta}$ and assuming that the interstellar extinction applies in the same way for the stellar population and the ionized gas in emission-lines regions.

A color-color diagram with the dereddened color indices (B - V)₀ and (V - R)₀ is shown in Fig. 7. The extranuclear HII regions and starburst nuclei are well mixed in this diagram, indicating identical stellar populations born during the same star formation episode. We first compare the colors of these starburst regions to the total color indices of "normal'' galaxies. It comes as no surprise that they are much bluer than quiescent galaxies whose position is indicated by the dotted rectangle in Fig. 7; the vast majority of our starburst regions are located outside this rectangle traced by about 500 normal galaxies (Buta & Williams 1995). We then use the predictions of stellar population synthesis models of Leitherer & Heckman (1995) to estimate the age of the stellar population which dominates the spectral continuum of these starburst regions. It appears clearly that the colors observed in the star-forming regions are well fitted by a very young stellar population with an age lower than 50 Myr.

$\begin{figure} \vspace{-0.5cm} \epsfxsize=8.8cm \centering{\mbox{ \epsfbox {ds1465f7.eps} }} \vspace{-0.5cm}\end{figure}$

Figure 7: Dereddened color-color diagram. The symbols have the same meaning as in Fig. 1. The dotted rectangle indicates the position of a sample of 500 normal galaxies (Buta & Williams 1995). The predictions of stellar population synthesis models of Leitherer & Heckman (1995) are indicated by the big stars for 5, 50 and 500 Myr

5.3 Reddening

The amount of reddening has not been estimated for 13 nuclear and 35 extranuclear emission-line regions because of the weakness or absence of H $\beta$ emission in their spectra. In all these objects we detect a relatively strong H $\alpha$ emission; they are thus probably highly obscured. Note that in a few objects (nucleus of Mrk 52, extranuclear regions of Mrk 489 and 712), the observed Balmer decrement is significantly less than the theoretical value; we assigned an internal extinction of zero to these objects.

$\begin{figure*} \epsfxsize=16cm \centering{\mbox{ \epsfbox {ds1465f8.eps} }} \vspace{-6.5cm}\end{figure*}$

Figure 8: Distribution of a) the extinction coefficient $c_{\rm H\beta}$ , b) the electron density $n_{\rm e}$ and c) the excitation parameter [OIII] $\lambda$ 5007/H $\beta$ for all the emission-lines regions (top), for the starburst nuclei (SBN) and the extranuclear HII regions (HII) (bottom)

5.4 Electron density

We derived the electron density ( $n_{\rm e}$ ) from the reddening-corrected [SII] $\lambda$ 6716/[SII] $\lambda$ 6731 line ratio using the analytical relation given by Osterbrock (1989). A 20% uncertainty in the [SII] $\lambda$ 6716/[SII] $\lambda$ 6731 flux ratio corresponds to an uncertainty of about 100 cm^-3 in the determination of $n_{\rm e}$ .

Figure 8b shows the distribution of the electron densities for the different emission-line regions. The mean value of the electron density is nearly the same for nuclear starbursts (560 $\pm$ 240 cm^-3) and extranuclear HII regions (770 $\pm$ 500 cm^-3). These mean values are higher than those derived for nearby HII nuclei (180 $\pm$ 200 cm^-3, Ho et al. 1997a), disk HII regions ( $\sim$ 140 cm^-3, Kennicutt et al. 1989) and luminous infrared starburst galaxies ( $\sim$ 280 cm^-3, Veilleux et al.1995).

5.5 Line-intensity ratios

5.5.1 [OIII]/H $\beta$ as an excitation parameter

Two types of starburst galaxies can be distinguished based on their level of excitation: SBNGs show low-excitation spectra ([OIII]/H $\beta$ < 3, see Fig. 2) whereas HII galaxies show high-excitation spectra ([OIII]/H $\beta$ $\geq$ 3, see Fig. 3).

A quick inspection of Fig. 1 indicates that the upper left region of the diagnostic diagrams contains only a few data points, reflecting the fact that essentially all the emission-line regions classified as starbursts have a relatively low excitation level ([OIII]/H $\beta$ $\leq$ 3). This is to be contrasted with Figs. 1-3 of Veilleux & Osterbrock (1987) where this region of the diagrams is populated with extranuclear HII regions and the low-metallicity HII galaxies from the sample of French (1980).

The distribution of the excitation parameter is shown in Fig. 8c. Our sample is obviously deficient in HII galaxies, since only five starburst galaxies (Mrk 86, 412, 803, 860 and 1346) have an excitation parameter [OIII]/H $\beta$ $\geq$ 3. The properties of these galaxies (listed in Table 1) indicate that they are mainly small, irregular and low-mass galaxies with a low dust content, confirming the general trend of this class of starburst galaxies (Coziol 1996). Our sample thus contains a vast majority of starburst nuclei located in more massive and chemically evolved galaxies than HII galaxies because of their higher frequency of past bursts of star formation (Coziol 1996). Note however that SBNGs are still in the process of formation because of their lower metal content compared to "normal'' spiral galaxies (Coziol et al.1997a). Figure 8c also shows that the mean excitation parameter is slightly higher in extranuclear HII regions ([OIII]/H $\beta$ $\sim$ 0.72) than in starburst nuclei ([OIII]/H $\beta$ $\sim$ 0.50), reflecting the negative abundance gradient from the nucleus to the outer parts of spiral galaxies.

$\begin{figure} \vspace{-0.5cm} \epsfxsize=8.8cm \centering{\mbox{ \epsfbox {ds1465f9.eps} }} \vspace{-0.5cm}\end{figure}$

Figure 9: Diagram of [SII]/H $\alpha$ as a function of [NII]/H $\alpha$ . The symbols have the same meaning as in Fig. 1. The mean value of [NII]/H $\alpha$ is higher in starburst nuclei (dotted-dashed line) than in extranuclear HII region (dashed line) whereas the mean values of [SII]/H $\alpha$ are identical (dotted line)

Three Wolf-Rayet galaxies are present in our sample. The optical spectrum of this subset of starburst galaxies shows broad emission lines from Wolf-Rayet stars around $\sim$ 4700 Å, the brightest line being HeII $\lambda$ 4686. While Mrk 52 and Mrk 710 were already known Wolf-Rayet galaxies and included in the catalog of Conti (1991), a new one, Mrk 712, was discovered in the sample (Paper I).

5.5.2 Excess of [NII]/H $\alpha$ in starburst nuclei

In the diagnostic diagram of Fig. 1a, one can see that, for a given excitation parameter, our nuclear starbursts tend to have stronger [NII] $\lambda$ 6583 emission than extranuclear HII regions, but the difference is rather small (< 0.1 dex) compared to other samples of starburst nuclei cited above.

This excess of nitrogen emission was first noted by Stauffer (1982) in the nuclei of "normal'' galaxies and confirmed later in a sample of "HII region-like'' nuclei by Kennicutt et al. (1989) who proposed the presence of a hidden weak AGN to account for this excess of low-ionisation emission line. AGNs indeed produce a harder ionizing radiation field than young O- or B-type stars. These high-energy photons create an extensive partially ionized zone from which low-ionization emission lines, such as [NII], [SII] and [OI] originate.

To be sure that no hidden AGN is located in the starburst nuclei of our sample, we compare in Fig. 9 two ratios of low-excitation emission lines, that of [SII]/H $\alpha$ and that of [NII]/H $\alpha$ . In the presence of a harder ionizing spectrum, both ratios should increase and a correlation would appear. One can clearly see that there is no such relation between the two ratios, neither for the nuclear starbursts nor for the extranuclear ones, which are well mixed in this diagram. In fact, the mean value of [SII]/H $\alpha$ is nearly identical for starburst nuclei ( $\sim$ 0.20) and extranuclear HII regions ( $\sim$ 0.21). The presence of a weak AGN in the nuclei of our starburst galaxies is also excluded because of the weakness of [OI] $\lambda$ 6300 in their spectrum: only $\sim$ 25% of our nuclear spectra show this emission line with intensities similar to those observed in normal HII regions (Veilleux & Osterbrock 1987).

Alternative explanations, like collisional excitation by shocks (Kennicutt et al. 1989) or very hot O type stars in metal-rich environments (Filippenko & Terlevich 1992; Shields 1992), have been suggested as ionization sources to account for the excess of nitrogen emission in galactic nuclei. However, both suggestions fail to reproduce our observations, because they also imply an increase of other low-ionization emission lines like [SII] and [OI].

We have investigated whether the impact of dust on the thermal properties of HII regions would provide a better explanation. The dust content in our sample is not negligible since all our galaxies are IRAS sources (one of our selection criteria). Calculations by Shields & Kennicutt (1995) indicate that the influence of dust on the emergent optical spectrum of HII regions can be quite appreciable in high-metallicity ( $Z \gt Z_{\odot}$ ) environments, as is the case in many galactic nuclei. In Fig. 1, we compare our data to the results of the photoionization model of Shields & Kennicutt (1995) which incorporates the effects of dust and is calculated for a stellar effective temperature of 45 000 K. The predicted line strengths do not provide a good match for all the observations in our starburst nuclei. The model accounts reasonably well for the [NII]/H $\alpha$ ratios observed in high metallicity ([OIII]/H $\beta$ $\leq$ 0.5) nuclei and for regions very close to the transition limit between HII and LINERs, but this appears to be accidental, since the predicted [SII]/H $\alpha$ ratio does not match our observations. These theoretical results might simply be the consequence of the selective initial element abundances, since Shields & Kennicutt (1995) arbitrarily assumed an enhancement of nitrogen abundance, with a secondary component scaling as Z², while other elements are in solar proportions. The results of the photoionization model of Shields & Kennicutt (1995) might thus follow from this selective abondance introduced ad hoc in the model.

$\begin{figure*} \epsfxsize=16cm \centering{\mbox{ \epsfbox {ds1465f10.eps} }} \vspace{-0.5cm}\end{figure*}$

Figure 10: Distribution of a) morphological type from RC3, b) inclination, c) heliocentric radial velocity, d) distance, e) absolute blue magnitude and f) apparent blue magnitude for SBNGs (solid line) and AGNs (dotted line). Mean values are indicated by vertical arrows

An enhancement of nitrogen in starburst nuclei is certainly the most reliable explanation to account for the moderate excess of nitrogen emission in our sample of galaxies. Such selective chemical enrichment of nitrogen has been observed in the interstellar medium of some nearby starburst galaxies, like NGC 5253 (Walsh & Roy 1989; Kobulnicky et al.1997) where N-enriched regions are found in the vicinity of young starbursts with a large population of massive Wolf-Rayet stars (Schaerer et al.1997). Moreover, chemical evolution models of galaxies (e.g. Marconi et al. 1994) predict an enhancement of nitrogen abundance after a succession of short and intense bursts of star formation, which has certainly been the case in starburst nuclei (Coziol 1996).