Category Archives: North Africa

Andalusian mtDNA highlights W-E differences

The issue of W-E genetic differences in Iberia has been discussed before in this blog by guest author Argiedude on the grounds of Y-DNA, showing that roughly the Western third of Iberia is distinct from the rest, most notably because of its higher presence of North African lineage E1b-M81. However the differences also appear in the mtDNA, even if they may be a bit more subtle because they cross with a N-S gradient of sorts.
A new study focused on the matrilineages of two characteristic Andalusian provinces, Granada in the East and Huelva in the West, underscores that this difference is very real.
Candela L. Hernández et al., Human maternal heritage in Andalusia (Spain): its composition reveals high internal complexity and distinctive influences of mtDNA haplogroups U6 and L in the western and eastern side of region. BMC Genetics 2014. Open accessLINK [doi:10.1186/1471-2156-15-11]

Abstract (provisional)


The archeology and history of the ancient Mediterranean have shown that this sea has been a permeable obstacle to human migration. Multiple cultural exchanges around the Mediterranean have taken place with presumably population admixtures. A gravitational territory of those migrations has been the Iberian Peninsula. Here we present a comprehensive analysis of the maternal gene pool, by means of control region sequencing and PCR-RFLP typing, of autochthonous Andalusians originating from the coastal provinces of Huelva and Granada, located respectively in the west and the east of the region.


The mtDNA haplogroup composition of these two southern Spanish populations has revealed a wide spectrum of haplogroups from different geographical origins. The registered frequencies of Eurasian markers, together with the high incidence and diversification of African maternal lineages (15% of the total mitochondrial variability) among Huelva Andalusians when compared to its eastwards relatives of Granada and other Iberian populations, constitute relevant findings unknown up-to-date on the characteristics of mtDNA within Andalusia that testifies a female population substructure. Therefore, Andalusia must not be considered a single, unique population.


The maternal legacy among Andalusians reflects distinctive local histories, pointing out the role of the westernmost territory of Peninsular Spain as a noticeable recipient of multiple and diverse human migrations. The obtained results underline the necessity of further research on genetic relationships in both sides of the western Mediterranean, using carefully collected samples from autochthonous individuals. Many studies have focused on recent North African gene flow towards Iberia, yet scientific attention should be now directed to thoroughly study the introduction of European genes in northwest Africa across the sea, in order to determine its magnitude, timescale and methods, and to compare them to those terrestrial movements from eastern Africa and southwestern Asia. 

Naturally the most noteworthy data is the frequency of mtDNA haplogroups in the two sampled populations:

A map comparing this data with some other areas of Iberia (mostly the West) is also provided:

Figure 3 – mtDNA haplogroup profiles registered in some populations of the Iberian Peninsula. The two Andalusian subpopulations studied here are marked with a red arrow. Codes are as in Additional file 3.
What makes Onubenses (the inhabitants of Huelva, ancient Onuba, West Andalusia) peculiar in relation to their Granadino neighbors is their lower frequency of H (notably H*, H3 and H5), along with their higher frequencies of K1, U3a and several North African related haplogroups (U6, L1b and L2). The peculiarities of Granadinos (notably the high H5 frequency) are not so notable in comparison.
Some of these Onubense peculiarities are reproduced in other parts of West Iberia, notably the low frequencies of H (but not at all in the NW corner), the presence of U6 (notably in North Portugal and also, not shown here, among the Maragatos of the León-Galicia border area) and to some extent the elevated frequency of K and some L(xM,N) lineages (varied localized frequencies).
Much of this seems best explained by ancient flows from NW Africa (flows which may be Neolithic, Paleolithic or from the Metal Ages but hardly related to Phoenician or Muslim colonization, which had no W-E gradient whatsoever) but I have some qualms about the quick identification by the authors of the origins of Onubense high K1 in that area. At the very least it must be noticed that, unlike the other African markers, K1 is much more common towards the Eastern parts of NW Africa and is also found at similarly high frequencies in many parts of Europe, such as France and Central Europe.
K1 was first spotted in Iberian ancient DNA in the early Neolithic (Los Cascajos, Navarre), and was an important Neolithic lineage through Europe. While I can’t discard the North African suggested origin, I think that other possibilities are at least as likely.
On U6, the presence of the very rare U6c, one of two basal lineages of U6, only found previously in 5 Moroccans, 10 Canarians and in one Italian, reinforces the idea of U6 expanding from West to East (against what most conclusions suggest on the most unclear grounds) with a most likely Moroccan origin. In turn this raises the question on how pre-U6 arrived to Morocco, especially as the Aurignacoid Dabban industries now seem to never have gone further West than Cyrenaica, opening the possibility of this lineage having arrived to NW Africa via Europe, where we know that U in general was common in the Upper Paleolithic, and which clearly influenced North Africa at the Oranian (aka Iberomaurusian) cultural genesis (LGM) via Morocco (Taforalt and other sites).
On the L(xM,N) lineages, it is worth mentioning that Cerezo 2012 claimed that some of them may be pre-Neolithic in Europe, especially L1b1a variants, which are widespread at low frequencies through Europe with an Iberian centrality.
The overall picture is maybe best visualized by the Hierarchical Cluster Analysis:

Figure 5 – Hierarchical Cluster Analysis (HCA) of 53 populations based on their mtDNAdiversity. The haplogroups used here are marked with arrows (vectors). Populations are indicated with numbers as in Additional file 3.
Cluster 1 is particularly noticeable for its high frequency of mtDNA H, being composed by mostly Iberian samples (along with South Germans and some Sicilians). Granada, as well as nearby Córdoba, sit here (even if with a tendency towards the more mainstream European Cluster 2). Huelva is not too remarkable when compared with other European samples in the mtDNA HCA, although it does show some deviation towards NW Africa, clustering closest to Canarians.
In spite of the good frequency of North African samples, it must be noted that the horizontal axis, in essence contrasting Europe vs. West Asia, has a weight of almost 50%, while the vertical axis, contrasting these two vs. North Africans (and the Sámi) only weights 13% (axis are almost never of the same relevance in PCA-like graphs, even if they are presented as such).
In synthesis: NW Africa had some minor but significant genetic influence in the West Iberian third, long before the Muslim period, and this mtDNA study ratifies these distinctions in the particular case of Andalusia, adding some rich detail of data.
Also, as the authors underline in their discussion, the issue of European genetic influence in North Africa still requires some serious investigation.

Hunter-gatherers, acorns and tooth trouble

It has been commonplace to believe that hunter-gatherers had good tooth health and that it was farming what caused dental problems because as cereals became a staple. There was good reason for that: caries were detected only rarely among hunter-gatherer remains (0-14%) while early farmers had much such painful problems much more frequently.
However the Upper Paleolithic people of Taforalt caves (Rif, North Africa), some 14,000 years ago (Oranian culture), had caries in 51% of adult teeth, a frequency comparable to those of early farmers.
This is attributed to the very high levels of nut consumptions, particularly acorns but also pine nuts, juniper berries, pistachios and wild oats. The number of acorn remains found is so large that the archaeologists had to conclude that they were used as year-long staple.

Late Upper Paleolithic of North Africa
· Iberomaurusian, aka Oranian, is shaded in dark green ·
· The core area of Capsian is shaded in gray-blue ·
(credit: Locutus Borg (anticopyright))
Taforalt people had hand mills, which they used to process some of these nuts, most likely the acorns, whose consumption as bread has been documented since antiquity.
Another finding are esparto grasses, which the authors believe were used in basketry. However I must mention that this versatile fiber has known many uses, being documented in Neolithic clothing of nearby Andalusia, used for some types of shoes even today and, of course, being a prime material for rope-making.

Esparto bale
Oranian culture dates to c. 22,000 years ago, with likely (partial?) roots in the Southern Iberian Gravetto-Solutrean (hence the name Iberomaurusian, although the culture as such is not known in Iberia). It was replaced in the Epipaleolithic by Capsian culture, with ultimate roots at the Nile (and hence the most likely vector of Afroasiatic languages leading to Tamazigh, aka Berber).
Source: PhysOrg.
Ref. Louise T. Humphrey et al., Earliest evidence for caries and exploitation of starchy plant foods in Pleistocene hunter-gatherers from Morocco. PNAS 2013 (pay per view, free after 6 months) → LINK [doi: 10.1073/pnas.1318176111]

Posted by on January 9, 2014 in Africa, Morocco, North Africa, Oranian, Paleolithic food


A western riverine route for human migration to North Africa in the Abbassia Pluvial

Interesting study on paleo-rivers of the Sahara providing insight for a likely route for Homo sapiens to cross the Sahara towards NW Africa.
Tom J. Coulthard et al., Were Rivers Flowing across the Sahara During the Last Interglacial? Implications for Human Migration through Africa. PLoS ONE 2013. Open accessLINK [doi:10.1371/journal.pone.0074834]


Human migration north through Africa is contentious. This paper uses a novel palaeohydrological and hydraulic modelling approach to test the hypothesis that under wetter climates c.100,000 years ago major river systems ran north across the Sahara to the Mediterranean, creating viable migration routes. We confirm that three of these now buried palaeo river systems could have been active at the key time of human migration across the Sahara. Unexpectedly, it is the most western of these three rivers, the Irharhar river, that represents the most likely route for human migration. The Irharhar river flows directly south to north, uniquely linking the mountain areas experiencing monsoon climates at these times to temperate Mediterranean environments where food and resources would have been abundant. The findings have major implications for our understanding of how humans migrated north through Africa, for the first time providing a quantitative perspective on the probabilities that these routes were viable for human habitation at these times.

Figure 2. Simulated probability of surface water during the last interglacial.
figure details Archaeological sites, and an annual probability that a
location has surface water. The archaeological data are derived from a
number of sources (including [42], [66], [67], [68].
The findspots are characterised by Aterian and Middle Stone Age
artefacts such as bifacial foliates and stemmed Aterian points and/or
typical ‘Mousterian’ points, side scrapers and Levallois technology.
Most are represented by surface scatters but where stratified examples
exist these can be shown by dating (OSL and U-series techniques) and
geomorphological setting to belong within MIS 5e [41], [42].

As discussed in other occasions, it seems likely that some genetic remnants of those early migrations are still visible in at least some NW Africans.

See also:


Algerian haploid genetics

This new study has particular interest for data miners willing to dig in the supplemental materials. It also has some other points of interest that I will discuss below and its general approach is loosely alright. However there are many nuances to be discussed in depth on the very complex NW African genetic landscape in which their tentative conclusions seem to lack enough depth of analysis (who grabs too much, squeezes little). Hence the complexity is too big for me to go issue by issue offering a criticism, so I will leave most of that open for the discussion, if the readers wish so.

Asmadan Bekada et al., Introducing the Algerian Mitochondrial DNA and Y-Chromosome Profiles into the North African Landscape. PLoS ONE 2013. Open accessLINK [doi:10.1371/journal.pone.0056775]
Mitochondrial DNA
The mtDNA landscape of Algeria and Northwest Africa is dominated (using HVS-I only to estimate it) by R-CRS (“H/HV” in table S2) with levels of 18-34% (29% in Algeria) almost comparable to Western Europe (~45%). This fraction we know from previous studies to be composed almost only by H1, H3, H4 and H7, all them attributed by Cherni to be originated (judging on diversity) in SW Europe (Iberia, France). Along with them HV0/V (7% in Algeria, 5-9% regionally) must be mentioned as also plausibly to be from that part of Europe (4-7%).
Another notable lineage is U6 (typical and most diverse in NW Africa), which reaches frequencies of 11% in Algeria (somewhat less in neighboring countries). Outside this area is only notable in Levant (~1%) and Iberia (~1,4%).
M1 reaching 7% in Algeria (~1-4% elsewhere in NW Africa, <1% in Europe and Highland West Asia, 1.2% in Levant, 2.4% in Peninsular Arabia) is also very much worth a mention, especially because the authors find an specifically NW African node centered in Algeria (HT2):

Figure 3. Reduced median network relating HVS-1 sequences of subhaplogroup M1.
(…) Black circles correspond to haplotypes observed in Algeria, whereas grey triangles pentagons correspond to lineages found in Egypt. Haplotype observed both in Algeria and Egypt are indicated using a black triangle. Grey circles indicate haplotypes observed in other geographical regions. (…)
The pattern suggests an Egypt-centered expansion for this lineage, however notice that East African M1 was not considered. 
Synthesis of mtDNA haplogroups or paragroups found in NW Africa at frequencies >2.5% (see table S2 for details and the many low frequency lineages as well), nomenclature as in table S2 (but some annotations in [square brackets] by me), frequencies for Algeria first (in brackets NW African range):
  • HV/H[R-CRS]: 28.8% (17.9-34.2%)
  • HV0/HV0a/V: 6.7% (4.6-8.3%)
  • R0a: 0.8% (0.8-3.2%)
  • U3*: 3.2% (1.1-3.2%)
  • U6a[U6a*]: 1.9% (1.9-7.8%)
  • U6a1’2’3: 9.4% (2.6-9.4%)
  • K*: 1.6% (0.7-4.8%)
  • T1a: 3.5% (0.0-5.6%)
  • T2b*: 1.9% (0.0-2.2%)
  • J[*]/J1c/J2[*]: 3.8% (1.3-3.8%)
  • M1[*]: 7.3% (0.7-7.3%)
  • L3b[*]: 0.3% (0.3-2.8%)
  • L3b1a3: 1.3% (0.0-2.8%)
  • L3e5: 1.6% (0.0-2.9%)
  • L2*: 0.5% (0.0-4.1%)
  • L2a[*]: 0.8% (0.0-3.2%)
  • L2a1*: 1.3% (0.7-4.8%)
  • L2a1b: 1.3% (0.8-3.5%)
  • L2d: 0.0% (0.0-2.8%)
  • L1b*: 3.0% (2.7%-9.0%)
Notice that in nearly all cases L(xM,N) highest frequency correspond to West Sahara. The exceptions are L2a* (Tunsian “Andalusians”) and L3e5 (Tunisians), suggesting maybe a local NW African deep rooting rather than ancient or recent flows from Tropical Africa. There are other lineages in the low frequency range in similar situation.
For this and other reasons I decided to color-code the list above according to my best guess about the origin of each lineage: NW African in deep red, Tropical African in brown, Egyptian in light brown, West Asian in green and European in blue. Unclear cases I left in black type.
Y chromosome DNA
Algerian and NW African Y-DNA is overwhelmingly dominated by E1b1b1b (M81), reaching 44% in Algeria (44-67% in the region), which is a NW African specific lineage. The second most important lineage by frequency is J1 (M304) with 22% in Algeria (0-22% in the region, 6-22% if we exclude Libya). None of the rest of the lineages reaches 7%, excepted E1b1b1c (M123) but only in West Sahara (11%, elsewhere it is very minor).
List of Y-DNA haplo-/paragroups with frequencies above 2.5% anywhere in NW Africa follows (based on table S6). Same notation as with mtDNA (Algerian frequency first, NW African range in brackets):
  • E1a (M33): 0.6% (0.0-5.3%)
  • E1b1[*] (P2): 5.2% (0.7-38.6%)
  • E1b1b1[*] (M35): 0.6% (0.0-4.2%)
  • E1b1b1a4 (V65): 1.9% (0.0-4.8%)
  • E1b1b1b (M81): 44.2% (44.2-67.4%)
  • E1b1b1c (M123): 1.3% (0.0-11.1%)
  • F[*] (M89): 3.9% (0.0-3.9%)
  • J1 (M267): 21.8% (0.0-21.8%)
  • J2a2 (M67): 3.9% (0.0-3.9%)
  • R1b1a (V88): 2.6% (0.9-6.9%)
  • R1b1b1a1b[*] (U198): 2.6% (0.0-2.6%)
  • R1b1b1a1b1 (U152): 2.6% (0.0-2.6%)
For more diverse samples of NW African Y-DNA (from previous studies), Wikipedia has a nice table.
I would like to highlight the problematic of J1 in Africa in general (including NW Africa). While there is no reasonable doubt that J1 as a whole originated in West Asia, it is found at rather high frequencies in East/NE Africa (Sudan, the Horn, Upper Egypt) and NW Africa with only very limited (at best) company by J2. Instead West Asian populations show a much more balanced apportion of the two major J sublineages, even in Saudi Arabia the J1:J2 proportion is of 8:3, almost 2:1. We do see this kind of apportioning in Lower Egypt, suggesting a “recent” (Neolithic or later) demic colonization from West Asia but we see exactly but nowhere else in Africa, where J1 is found always much more frequently than J2 (if the latter is found at all). 
In my understanding this excludes colonization from West Asia after the Pre-Pottery Neolithic B, which seems the most plausible scenario for the spread of “Highlander” J2 into “Lowland” West Asia (probably dominated by J1 initially). So J1 in Africa (excepted Lower Egypt) cannot be argued easily to be of “recent” Neolithic, much less Semitic or Arab origin: it must be older. 
Also Ethio Helix commented in this very interesting discussion at his blog that Tofanelli 2009 found low diversity on NW African J1. However, to my knowledge, nobody has looked at NE/East African J1 diversity nor a proper study has been done on the substructure of this lineage in Africa. This leaves wide open the possibility that NW African J1 has a NE African origin, surely related to the expansion of Capsian culture or internal African Neolithic flows. 
While this matter is not properly addressed, researchers will oversimplify and imagine J1 as simply West Asian influx. It is ultimately of course but I strongly suspect that it has a secondary and distinct NE African center at the Nile basin and this is being totally ignored. 
This study offers several rough comparisons with nearby regions (but not West Africa), however they oversimplify some stuff (the already mentioned Y-DNA J1 or assigning all mtDNA L(xM,N) to East Africa, when it seems obvious that some lineages may be deeply rooted in NW Africa or others probably come from West Africa). For whatever it is worth anyhow, here there are two such questionable comparisons:

Table 2. Geographic components (%) considered in Y-chromosome and mtDNA lineages.

Figure 2. Graphical relationships among the studied populations.
PCA plots based on mtDNA (a) and Y-chromosome (b) polymorphism. Codes are as in Supplementary Tables S2 and S6.

See also:


Posted by on February 23, 2013 in African genetics, mtDNA, North Africa, West Eurasia, Y-DNA


Evidence of marine exploitation 250,000 years ago in North Africa

Dr. Cantillo in a cave access
According to news reports, Juan Jesús Cantillo the University of Cádiz has argued in his (successful) doctoral thesis that the exploitation of marine resources in Benzú Cave (Ceuta, North Africa) has some 250,000 years of antiquity instead of the mere 100,000 that has been proposed for such kind of economy by other scholars always in search of absolutist dividing lines between what is “modern human” and what is something else. 
99% of the coastal resources exploited by the ancient inhabitants of Benzú are limpets, albeit of a variant quite larger than modern ones. While no bones have been found that could inform us of the human species involved in this economy of coastal exploitation, some artifacts appear to be similar to those used by Neanderthals across the Gibraltar Strait. If confirmed, this would also imply intercontinental navigation, even if across a narrow strait of maybe some 5 km (in the worst of the Ice Ages, today it has 14.3 km).
Source[es]: El Pueblo de Ceuta (h/t Pileta de Prehistoria). I could not find the thesis online yet but it says it was successfully defended earlier this month.

Posted by on December 16, 2012 in navigation, Neanderthal, North Africa


Mitochondrial haplogroups M1 and U6

A new study has been published on the two most relevant African matrilineages of Eurasian origin: M1 and U6. There are others but these two are the ones, together with X1 maybe, who have a more typically African distribution with limited presence in Eurasia.
Erwan Pennarun et al., Divorcing the Late Upper Palaeolithic demographic histories of mtDNA haplogroups M1 and U6 in Africa. BMC Evolutionary Biology, 2012. Open accessLINK [doi:10.1186/1471-2148-12-234]
I must say that I am rather unpersuaded of what the authors have to say, very especially in regards to U6 (but not either on M1) but it is still a study with interesting data for the record.
Very briefly the authors use molecular-clock-o-logy (academic pseudoscience when used as alleged evidence of anything, being as it is mostly speculation with a mathematical pretext), along with very debatable archaeological interpretations, to propose that the expansion of U6 and M1 originate in the mid to late Upper Paleolithic and not, as Olivieri proposed in 2006 (and I rather support) from around the time of the (re-)colonization of West Eurasia c. 50-30 Ka ago.
Whatever the case they leave us with some data on these two lineages and some of their main subclades (not all). These can be found in table 1 and the supplementary materials. 
They also provided us with these frequency maps (enriched with approx. linguistic boundaries):

Absolute frequency of (A) M1 and (B) U6 (and approx. linguistic family boundaries)

Dabban industries
One of the contention points is the Dabban industries of Cyrenaica, which have been argued to be Aurignacoid and therefore correspond to the Homo sapiens (back-)migration into West Asia, Europe and probably also North Africa. The authors cite some references to question that the Dabban are derived from Palestinian technologies but I can only imagine (without getting into the matter in depth as of now) that the issue is under debate (as happens with other technologies from Europe, etc.)
A bigger problem is that no Dabban nor related industries are known to have existed in NW Africa, which is the region where U6 has its greatest basal diversity (clearly) and one of a few regions with greatest M1 basal diversity (the other two being Egypt and Arabia). 
U6 from NW Africa or what?
This is another element of the paper that irks me: that the authors happily reject the weight of basal diversity, which is clearly in NW Africa (Morocco notably) and around it (Canary Islands, Iberia). Yet the authors contend:

Whilst several U6 sub-clades seem to be confined to Northwest Africa, this pattern may be the result of drift and founder effects over many thousands of years and does not necessarily suggest that Northwest Africa was the geographic source of U6 dispersals in Africa.

The reasoning is simply not sound, contradicting the logic of greatest parsimony. What they say might hypothetically have happened but has a low chance, notably in absence of any other evidence.
Also the authors argue, based mostly on their own age estimates (which, I insist, are speculations, educated guesses, and cannot ever be used as evidence) that U6 may have arrived with Oranian (aka Iberomaurusian), of quite likely Gravetto-Solutrean South Iberian relatedness. This Western European connection is not only supported by some quite reasonable typological likenesses and mutual influences (back-tipping weapon points in Iberia may have an Aterian origin for example) and chronology of the sites (generally older towards the West) but also by anthropometric considerations (many of the best known Crô-Magnon type specimens, related to Gravettian industry in Europe, are North African Oranian) but also the fact that North Africa has loads of mtDNA of likely West European origin¹ (H1, H3, H4, H7 and V, amounting together to some 30% of the lineages of the region) and that Oranian ancient mtDNA was found to be consistent with this kind of mtDNA pool by Kefi in 2005² (however, as only HVS-I was tested for, the certainty is not 100% – but not your usual African L(xM,N) in any case).
So even if U6 would be of Oranian origin, it would still most likely have coalesced in Morocco (or somewhere nearby), something the authors seem reluctant to admit without offering any clear alternative.
And M1?
M1 has three regions of high basal diversity (only two subclades exist however M1a and M1b, making this kind of evaluation a bit less certain); they are: Arabia, Egypt and NW Africa.
The closest relative of this haplogroup is M20’51, which is found in SE Asia (Vietnam, South China, Indonesia, etc.) as well as Nepal. I do use this kind of “sister” references also to estimate the most likely origin jointly with basal diversity, and, in this case it suggests that M1 is from Arabia and that North African (and also East African and Highland West Asian) M1 is derived from this origin.
The authors don’t seem to take a clear stand in this case either and they even do not mention the East Asian relative, insisting on analyzing only their molecular-clock-o-logic speculations, what are at best only marginally relevant.
However the concede that M1 and U6 were in Africa long before the Afroasiatic expansion.
In brief
In the end a somewhat messy study with too much emphasis in the wrong stuff but still good for the data. Check, please table 1 and, if you feel like researching the matter in greater depth the supplemental materials, which are no doubt informative in their own right.


¹ See these papers:

  • H. Enafaa, V. M. Cabrera et al., Mitochondrial DNA haplogroup H structure in North Africa. BMC Genetics 2009. Open Access. [LINK]
  • L. Cherni et al., Post-last glacial maximum expansion from Iberia to
    North Africa revealed by fine characterization of mtDNA H haplogroup in
    . American Journal of Physical Anthropology. Pay per view. [LINK]
  • Claudio Ottoni et al., Mitochondrial Haplogroup H1 in North Africa: An Early Holocene Arrival from Iberia. PLoS ONE 2010. Open Access. [LINK]

² R. Kéfi et al., Diversité mitochondriale de la population de Taforalt
(12.000 ans bp – maroc): une approche génétique a l’étude du peuplement
de l’afrique du nord
. Anthropologie 2005. [PPT presentation direct download – Institut Pasteur]


Visual etymological

Spanish archaeological blog Asociación los Dólmenes reports today[es] that a curious and somewhat obscene finding is at the roots of the modern city of Seville (known as Hispalis in Roman times). The finding of a phallic relief on the entrance of one of the oldest buildings of that city, at the port, has open a debate on whether the city has its origins in whore house (as could be normal for a harbor) or are we talking instead of a building-protector deity apparently of North African origins (where is found in many public buildings).

But regardless of the exact meaning of the icon, the depiction of an erect virile member with avian legs made me think of the origin of colloquial Spanish and English words for penis: cock and polla (Sp. chicken, fem.) Obviously Romans were not thinking of T. rex, right?
What about other languages? Berber, Portuguese, Catalan-Occitan, French, Italian? 
PS: The image actually has a lizard-like tail what should get us all a bit perplex because the closest thing that comes to mind is a dinosaur but Romans could not know anything about dinos, could they? 
This is the kind of argument used to reject the authenticity of some archeological findings like in the Iruña-Veleia case, where conjectures about the plausibility or not of this or that text (the non-existent Descartes – is Miscart) or letter (Z for example) have been used as alleged proof of falsification
Whatever the deep logic behind this icon, it’s not weirder than gargoyles or centaurs, is it?