Introduction

The embodiment of communication is attracting an increased interest. This interest was sparked by hypotheses and findings surrounding the so-called “mirror neurons” (Arbib 2005), neurons that are active during the production as well as the perception of actions. Neurons with these properties have been observed in area F5 of the macaque monkey (Gallese et al. 1996; Kohler et al. 2002; Rizzolatti et al. 1996). Similar properties have been detected within inferior frontal gyrus (IFG), also referred to as Broca’s area in humans (Buccino et al. 2001; Iacoboni et al. 1999; Rizzolatti and Arbib 1998; Rizzolatti and Craighero 2004). Broca’s area in turn has been associated with speech generation. In his now classic report, Pierre Paul Broca (1861) described a man with a cavity in his left frontal lobe who was unable to speak fluently, leading to the notion that Broca’s area is involved in speech production. Furthermore, Broca’s area has been implicated in spoken language comprehension (Gough et al. 2005; Moss et al. 2005).

Recently a growing number of studies have proposed a link between IFG involvement in imitation and its involvement in speech (Arbib 2005; Hamzei et al. 2003; Kühn and Brass 2008; Nishitani et al. 2005; Watkins and Paus 2004). These studies implicitly assume that the anatomical association between regions involved in imitation and language processing ought to occur for non-arbitrary reasons. This link could be explained by the strong relationship between hand gestures and speech. Oftentimes gestures are displayed even though the gesturer is aware of the fact that the person he is talking to cannot see his gestures, e. g., when using a telephone. Congenitally blind individuals have been shown to gesture even when speaking to other blind people who likewise cannot perceive the hand movements (Iverson and Goldin-Meadow 1998). Furthermore, the link between manual gestures and speech is supported by stutterers whose speech-related gestures freeze when their speech stops, whereas unrelated hand movements usually continue (Mayberry and Jacques 2000). In a neuroimaging study on sign language, the associated observed brain activation was similar to activation during overt speech (Levänen et al. 2001). These findings suggest a co-representation of speech and gesture in Broca’s region and may reflect shared evolutionary roots of both functions.

A model on the stages of language’s phylogeny by Arbib (2005) attempts to bridge the explanatory gap between mirror neurons, imitation, gestures and speech. He claims that language and gesture developed in an expanding spiral in which one component prepared the ground for further development of the other component, so language scaffolded gesture just as gesture did for language. Assuming that these evolutionary connections exist, one might expect to find rudiments of them in those functions at present, e. g., in terms of overlapping brain activation in neuroimaging studies on imitation and speech.

In order to explore the overlap of neural correlates of imitation and speech, the present study employed quantitative meta-analyses methods to assess the correspondence of neural activations across multiple neuroimaging studies on imitation and speech using the activation likelihood estimation (ALE) approach (Eickhoff et al. 2009; Laird et al. 2005; Turkeltaub et al. 2002). This approach reveals statistically significant concordance of activated voxels across multiple studies controlling for chance clustering. By seeking concordance at the voxel level, ALE tests for statistically reliable clustering of activations in standardized locations, avoiding spatial distinction errors and problematic incongruence of labeling across studies that can befall narrative-based reviews. With a subsequent contrast analysis, we then tested for unique locations of activity in imitation compared to speech within Broca’s area.

Materials and methods

Selection of studies

For the imitation meta-analysis, we selected studies involving hand imitation without the involvement of objects. The coordinates were taken from a recent ALE meta-analysis on action observation and imitation (Caspers et al. 2010).

For the speech meta-analysis we used the database Brainmap Sleuth (http://brainmap.org/sleuth/index.html). We used the search terms [Diagnosis = Normals] AND [Behavioral Domain = Cognition.Language.Speech]. From the resulting papers, we selected those that presented contrasts reflecting brain activity during speech generation in comparison to a control condition. In order to test whether ALE subtraction analysis can potentially reveal unique activation within Broca’s area, we searched for studies utilizing another task that reliably activates IFG but is not hypothesized to share a similar neural basis with imitation; we selected fMRI studies on inhibition which have reliably been associated with activity in right IFG (Aron and Poldrack 2005; Aron et al. 2004), including studies using the stop signal task and Go/NoGo task. We included coordinates resulting from analyses computed across the whole brain and not restricted using partial coverage, regions of interest or small volume correction.

We included data from fMRI and PET studies despite the fact that they have a different physiological basis because both methods have been used to identify the neural correlates of imitation and speech. Our rationale was to provide an all-embracing overview over the attempts to identify the neural correlates of imitation and speech. In total, a number of 20 object-free hand imitation studies with 247 foci of altogether 298 participants (Table 1), 58 overt speech studies with 1,401 foci of altogether 804 participants (Table 2) and 19 external inhibition studies with 123 foci of altogether 329 participants were included (Table 3).

Table 1 List of included studies on object-free hand imitation
Full size table
Table 2 List of included studies on speech generation
Full size table
Table 3 List of included studies on inhibition
Full size table

Creation of ALE maps

The ALE method provides a voxel-based meta-analytic technique for neuroimaging data (Eickhoff et al. 2009; Turkeltaub et al. 2002). By means of the software Brainmap GingerALE (http://brainmap.org/ale/), statistically significant concordance in the pattern of brain activity among several independent experiments was computed. ALE maps display regions in the brain that comprise statistically significant peak activation locations from multiple studies. Coordinates reported in Talairach were converted to MNI using Lancaster’s et al. (2007) transformation (icbm2tal). In the approach taken by ALE, localization probability distributions for the foci are modeled at the center of 3-D Gaussian functions, where the Gaussian distributions are summed across the experiments to generate a map of inter-study consistencies that estimate the likelihood of activation for each voxel, the ALE statistic, as determined by the entire set of studies. The false discovery rate (FDR) method was employed to correct for multiple comparisons at a significance threshold of p < 0.01 and a cluster threshold of 100.

Contrast analyses were calculated by means of ALE subtraction analysis, accounting for potential differences in sample size. To increase the specificity of the results, the analysis of differences was restricted to those voxels that showed an effect in main speech or imitation meta-analyses. The reported contrast analyses were thresholded at a corrected p value of <0.05.

Results

The results of the quantitative ALE analyses within IFG on imitation and speech as well as motor inhibition are presented in Fig. 1a–c.

Fig. 1
figure 1

ALE meta-analysis maps of bilateral IFG for neural correlates of object-free hand imitation (red) and a speech (green), b complex speech (cyan) or simple speech (yellow) and c motor inhibition (blue); p < 0.01, FDR corrected; the clusters shown are restricted to bilateral IFG defined by means of the Harvard Oxford Atlas

Full size image

In order to explore the similarity of brain activation in imitation and speech within Broca’s area, the first analysis aimed at comparing the ALE maps resulting from neuroimaging studies on object-free hand imitation and overt speech (Fig. 1a). Direct overlap was observed in the left hemisphere at coordinate −54, 12, 7. Since the included speech studies spanned a broad range of stimuli from simple syllables to free word generation that had to be uttered, we split them up into simple speech containing syllables and repeated sequences and complex speech containing words and sentences (Fig. 1b). Direct overlap was observed between imitation and complex speech at coordinate −54, 12, 5 in the left hemisphere and between imitation and simple speech at 59, 11, 11 in the right hemisphere. By means of contrast analyses, we investigated whether the concurrence observed within studies on object-free hand imitation was different to the concurrence within studies on simple as well as complex speech. In both subtraction analyses, no significant unique localization for imitation (nor speech) was found.

To provide evidence that unique location of activation within the region of IFG can technically be detected by means of ALE contrast analyses, we computed an ALE meta-analysis in a different behavioral domain, namely motor inhibition. Motor inhibition has typically been localized in IFG, in particular the right IFG (Aron and Poldrack 2005) (Fig. 1c). When performing a contrast analysis between imitation and inhibition, we found a selective cluster for imitation in right IFG: 60, 15, 12. Hence, we conclude that the observed overlap and, in particular, the absence of unique clusters for imitation and speech are meaningful and argue that the similarity of activation is not due to spatial constraints within Broca’s area.

Discussion

The present study aimed at exploring the proposed link between imitation and speech by means of comparing the neural correlates of both functions within Broca’s area employing quantitative meta-analyses tools. Assuming that an evolutionary connection between imitation and speech exists, we hypothesized to find rudiments of this link in terms of overlap in brain activation of both processes. In line with our predictions, we found overlap between the brain regions implicated by ALE meta-analysis on imitation and a separate meta-analysis on overt speech in left IFG.

The localization of concurrence for studies on object-free hand imitation in bilateral pars opercularis of IFG (BA 44) is in line with the results for imitation in the study of Caspers et al. (2010) that includes studies with imitation of other effectors such as foot or hand as well as object-related actions. We decided to only include manual object-free imitation since we considered this to be closest to typical speech-accompanying gestures. Our meta-analysis results (including 20 imitation studies) as well as the results of the previous study (including 32 imitation studies, Caspers et al. 2010) is at odds with a study by Molenberghs et al. (2009) that reported ALE concurrence only in dorsal premotor cortex (BA 6) not in the IFG (BA 44). Since the meta-analysis by Molenberghs and colleagues includes less studies that used a region of interest (ROI)-based approach and used a by now out-dated version of the ALE algorithm (Eickhoff et al. 2009), we feel confident to locate concurrence of object-free hand imitation in pars opercularis of the IFG. Likewise, the results of our overall speech analysis are in line with a previous meta-analysis on overt speech in non-stuttering subjects on eight studies (Brown et al. 2005) reporting a peak coordinate close to the one we found in left IFG. Contrary, another meta-analysis by Turkeltaub et al. (2002) on 11 studies reported a peak coordinate within the left precentral gyrus (BA4/6) not in left IFG. But the authors themselves have suggested that peaks within left IFG were likely smoothed over by the high ALE values in the adjacent motor strip.

In order to differentiate between speech studies with simple syllable utterances that are most likely rather melodic and those in which complex words or entire sentences had to be pronounced, we computed two separate meta-analyses. We found stronger activity in the right hemisphere during simple syllable production and a stronger left lateralization during more complex word and sentence production. This is in agreement with a previous studies showing that singing (Ozdemir et al. 2006) as well as prosodic modulation in speech perception (Meyer et al. 2002) is more strongly associated with right hemispheric activation, since the repetition of syllables likely results in utterances that bear resemblance with singing. Similarly, it has been reported that patients with Broca’s aphasia are able to sing the lyrics of a song better than they can speak the same words (Gerstman 1964; Yamadori et al. 1977) a finding that is also in line with a stronger left lateralization for speech as compared to melodic syllable utterances.

When comparing the localization of simple and complex speech with imitation-related concurrence, we found that complex speech-related activity overlapped with imitation within left IFG, whereas simple speech-related activity showed overlap within the right IFG. Subtraction analyses between simple and complex speech and imitation maps revealed no regions with unique localization, neither for speech, nor for imitation. This absence of distinct neural correlates within Broca’s area argues in favor of an association between imitation and speech functions.

To preclude that it is technically impossible to detect unique areas of activation within Broca’s area when computing a contrast of ALE analyses between coordinates of imitation and speech studies, we computed a meta-analysis on stop signal and Go/NoGo tasks that have commonly been associated with right IFG activity (Aron and Poldrack 2005; Aron et al. 2004) and compared it by means of an ALE contrast analysis with the imitation ALE. Interestingly, the inhibitory functions of right IFG have likewise been associated with the inhibition of imitative and overlearned responses (Brass et al. 2005). Therefore, the function of motor inhibition is not entirely unrelated to imitation. When subtracting the resulting regions of concurrence in motor inhibition processes from the imitation associated neural correlates, we found regions that are more strongly involved in imitation compared with motoric inhibition. The fact that motoric inhibition and imitation seem to have separable neural correlates underlines the importance of the observed absence of unique localization when comparing imitation and speech.

The observed overlap between neuroimaging correlates of object-free hand imitation and overt simple as well as complex speech is in line with the assumption that speech developed from gestures and manual actions (Arbib 2005). Furthermore, it fits to the motor theory of speech perception formulated by Liberman et al. (1952). According to this theory, the listener does not solve the invariance problem in speech perception in the auditory domain, but in the motor domain instead. Assuming that the acoustic patterns of speech can be different, the articulatory gestures that are needed to produce them are the same. Therefore, the idea is that the perceptual problem is solved by recruiting the production system of the listener. The listener mirrors the neuromotor commands of the articulators (e. g., tongue or lips) to understand the message (Liberman and Mattingly 1985). Further support for the intimate link of imitation and speech also comes from aphasic patients who are often apraxic, and apraxia often involves deficits in imitation (Heilman and Valenstein 2002). However, it is not yet clear whether this joint occurrence of syndromes results solely from a co-localization of praxis and speech or whether it can be regarded as a proof of its common mechanisms.

Defining the functional involvement of Broca’s region on a higher level, namely as representing sequential information could explain its importance for speech as well as imitation (Nishitani et al. 2005). Lesion data suggest that the two hemispheres might have different roles in sequencing. Left hemispheric lesions affect predominantly verbal sequencing, whereas right hemispheric lesions affect nonverbal sequencing (Bookheimer 2002).

Conclusions

Taken together, the current meta-analysis has shown considerable overlap and no unique localization for overt speech and object-free manual imitation-associated neural correlates within Broca’s area. This is necessary though not sufficient evidence for the notion that imitation and speech have a common neural basis and is in line with patient data and previous theoretical accounts on the evolutionary roots of speech in imitation. Therewith, the current meta-analysis provides first evidence for neuroanatomical overlap between speech and imitation that has previously only been inferred.